- Email: [email protected]
Shunt Failure—Risk Factors and Outcomes: An Analysis of The Society of Thoracic Surgeons Congenital Heart Surgery Database
RICHARD E. CLARK AWARD
Shunt Failure—Risk Factors and Outcomes: An Analysis of The Society of Thoracic Surgeons Congenital Heart Surgery Database Nhue Do, MD, Kevin D. Hill, MD, MS, Amelia S. Wallace, MS, Luca Vricella, MD, Duke Cameron, MD, James Quintessenza, MD, Neil Goldenberg, MD, PhD, Constantine Mavroudis, MD, Tom Karl, MD, Sara K. Pasquali, MD, MS, Jeffrey P. Jacobs, MD, and Marshall L. Jacobs, MD Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania; Duke Clinical Research Institute, Durham, North Carolina; Division of Cardiac Surgery, Massachusetts General Hospital, Boston, Massachusetts; Cincinnati Children’s Hospital, Cincinnati, Ohio; Johns Hopkins All Children’s Hospital, St. Petersburg, Florida; Florida Hospital, Orlando, Florida; and C.S. Mott Children’s Hospital, Ann Arbor, Michigan
Background. Systemic-to-pulmonary shunt failure is a potentially catastrophic complication. We analyzed a large multicenter clinical registry to describe the prevalence and evaluate risk factors. Methods. Infants (aged £365 days) undergoing shunt operations (systemic artery-to-pulmonary artery or systemic ventricle-to-pulmonary artery) in The Society of Thoracic Surgeons Congenital Heart Surgery Database (STS-CHSD) from 2010 to 2015 were included. Multivariable logistic regression was used to evaluate risk factors for in-hospital shunt failure. Model covariates included patient characteristics, preoperative factors, procedural factors including shunt type, and center effects. Centers with more than 15% missing data for key covariates were excluded. Results. Shunt operations were performed in 9,172 infants (118 centers). In-hospital shunt failure occurred in 674 (7.3%). In multivariable analysis, risk factors for in-hospital shunt failure included lower weight at operation (odds ratio [OR], 1.35; p [ 0.001), preoperative hypercoagulable state (OR, 2.47; p [ 0.031), and the
presence of any other STS-CHSD preoperative risk factors (OR, 1.24; p [ 0.038). Shunt failure was less likely with a systemic ventricle-to-pulmonary artery shunt than a systemic artery-to-pulmonary artery shunt (OR, 0.65; p [ 0.020). Neither cardiopulmonary bypass nor singleventricle diagnosis was a risk factor for shunt failure. Patients with in-hospital shunt failure had significantly higher rates of operative mortality (31.9% vs 11.1%, p < 0.001) and major morbidity (84.4% vs 29.4%, p < 0.001), and longer median postoperative length of stay among survivors (45 vs 22 days, p < 0.001). Conclusions. In-hospital shunt failure is common, and associated mortality risk is high. These data highlight at-risk patients and procedural cohorts that warrant expectant surveillance and may benefit from enhanced antithrombotic prophylaxis or other management strategies to reduce shunt failure. These findings may inform planning of future clinical trials.
S
these patients, creation of a systemic ventricle-topulmonary artery shunt is an alternative to the traditional systemic artery-to-pulmonary artery shunt [2, 3]. Early failure of both categories of palliative systemic-topulmonary shunts is an incompletely understood phenomenon that accounts for important morbidity and death among infants and neonates. Multicenter studies and clinical trials have focused primarily on pharmacologic strategies to reduce the risk of shunt failure caused by thrombosis [4–6]. Their utility for
hunt operations connecting the pulmonary arteries to a systemic artery or the systemic ventricle are common in the management of neonates and infants with congenital heart disease. Originally introduced to treat patients with cyanosis related to diminished pulmonary blood flow [1], systemic-to-pulmonary artery shunts are also used to regulate pulmonary blood flow, especially for patients undergoing palliation of single-ventricle anomalies such as hypoplastic left heart syndrome (HLHS). For
(Ann Thorac Surg 2017;-:-–-) Ó 2017 by The Society of Thoracic Surgeons
Accepted for publication June 6, 2017. Presented at the Fifty-third Annual Meeting of the Society of Thoracic Surgeons, Houston, TX, Jan 21–25, 2017. Winner of the Richard E. Clark Award for Congenital Heart Surgery. Address correspondence to Dr Jacobs, Division of Cardiac Surgery, The Johns Hopkins Hospital, Zayed 7107, 1800 Orleans St, Baltimore, MD 21287; email: [email protected].
Ó 2017 by The Society of Thoracic Surgeons Published by Elsevier Inc.
The Supplemental Tables and Figure can be viewed in the online version of this article [http://dx.doi.org/10. 1016/j.athoracsur.2017.06.028] on http://www.annals thoracicsurgery.org.
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2017.06.028
2
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
DO ET AL
guiding clinical decision making has been limited. Some have been underpowered, and some have had limited risk adjustment of patients. Many single-center studies have not considered the multifactorial nature of early shunt failure, focusing primarily on thrombosis [7, 8]. Using data from The Society of Thoracic Surgeons Congenital Heart Surgery Database (STS-CHSD), we sought to characterize the group of patients who were observed to have experienced early in-hospital shunt failure. This information should be helpful in risk assessment of individual patients and may enhance the precision and productivity of future prospective studies by informing eligibility criteria, risk stratification, and covariate adjustment.
Patients and Methods Data Source The STS-CHSD collects data on neonates, infants, children, and adults undergoing operations for congenital heart disease, currently including more than 400,000 operations conducted at 127 centers in North America. The database encompasses more than 95% of United States centers performing pediatric heart operations and approximately 98% of all pediatric cardiac operations done annually in the United States [9]. Coding is accomplished by clinicians and ancillary support staff using International Pediatric and Congenital Cardiac Code terminology, including demographics, anatomic diagnoses, noncardiac abnormalities, preoperative risk factors, intraoperative details, surgical procedures, postoperative complications, postoperative length of stay, and operative mortality. Data on all patients undergoing congenital heart operations at participating centers are entered into STS-CHSD data collection software. The Duke Clinical Research Institute is the data warehouse and analysis center for STS national databases. Evaluation of data quality includes intrinsic data verification and a formal process of site visits and data audits conducted by independent data quality personnel and pediatric cardiac surgeons at 10% of participating institutions annually. The STS-CHS Database Access and Publications Committee and the Duke University Institutional Review Board approved the study. The study was not considered human subjects research by the Duke University Institutional Review Board in accordance with the Common Rule (45 CFR 46.102[f]).
Ann Thorac Surg 2017;-:-–-
centers with more than 15% missing data for key covariates were excluded. Patients were considered to have experienced early shunt failure if one or more of three criteria were met: (1) diagnosis of “shunt failure” or (2) procedure code “shunt, reoperation” or (3) any “additional procedure” from a list of specific shunt interventions, entered in association with an event subsequent to the eligible index operation (Supplemental Table 2).
End Points The primary end point was ascertainment of early shunt failure, as defined above. Secondary end points were operative mortality, discharge mortality, and composite morbidity. Composite morbidity was defined as occurrence of any one or more of the following complications: renal failure requiring temporary or permanent dialysis, neurologic deficit persisting at discharge, arrhythmia necessitating permanent pacemaker, postoperative mechanical circulatory support, paralyzed diaphragm, and cardiac arrest. Definitions of the individual variables captured in the composite end point can be found in the STS CHSD specifications at the STS Web site [10].
Analysis Characteristics of patients, including distributions of demographic, preoperative, and operative factors, were explored and described using standard summary statistics. Pearson c2 tests for categoric variables and c2 rank-based group mean score statistics (equivalent to Wilcoxon tests) for continuous/ordinal variables were used to assess univariate differences in the distribution of patient characteristics across groups with early shunt failure and no shunt failure. Multivariable analysis was used to evaluate associations between potential risk factors and the occurrence of early shunt failure (as defined above). Conditional logistic regression stratified by center was used to account for center-level variability in practices. Potential risk factors for shunt failure included in the models as covariates were defined a priori based on the clinical judgement of the investigative team. Model covariates included patient characteristics in the categories listed above, in addition to the mutually exclusive shunt procedure types and diagnostic categories (HLHS, non-HLHS single ventricle, other). Results are summarized using estimated odds ratios (ORs), confidence intervals of the estimated ORs, and p values.
Patient Population
Sensitivity Analysis
Neonates and infants (aged 0 to 365 days) undergoing cardiac operations from January 1, 2010, through December 31, 2015, at STS-CHSD centers were potentially eligible for this study. The study population consisted of all patients for whom the index operation included one or more of the systemic-to-pulmonary shunt procedures included in STS CHSD data collection versions 3.0 or 3.22, or both (Supplemental Table 1). Hybrid Norwood procedures and combined Norwood and Rastelli (Yasui) procedures were excluded. Data from
Analyses were repeated after exclusion of all patients whose shunt type was a systemic ventricle-to-pulmonary artery shunt (Norwood with “Sano modification”; ie, ventricle-to-pulmonary artery conduit). This enabled us to estimate the prevalence of early shunt failure and assess associations of patient and procedural factors and outcome variables in a cohort limited to systemic arteryto-pulmonary artery shunt operations. Analyses were performed using SAS software (SAS Institute Inc, Cary, NC).
Ann Thorac Surg 2017;-:-–-
RICHARD E. CLARK AWARD DO ET AL EARLY SHUNT FAILURE
Results
3
with or without cardiopulmonary bypass support for the shunt procedure.
Study Population Characteristics A total of 9,172 patients from 118 institutions were eligible for analysis and considered to be potentially at risk for early shunt failure. In-hospital shunt failure was identified in 674 patients (7.3%) of the 9,172 who had undergone a shunt procedure as part of the index operation. Of the three study criteria for ascertainment of early shunt failure (Supplemental Table 2), diagnosis of “shunt failure” was coded in 267 patients; procedure “shunt, reoperation” was coded in 328, and the “additional procedure” criteria were met in 398. In 273 instances, the patient met two or all three of the criteria. The types of shunts performed and the occurrence of shunt failure for each type are reported in Table 1. The most commonly performed was the modified BlalockTaussig shunt (mBTS) in a setting other than a Norwood procedure, with 3,474 (37.9%) overall. The second most common was a systemic ventricle-to-pulmonary artery shunt (Sano) as part of a Norwood procedure. These outnumbered systemic artery-to-pulmonary artery shunts as part of a Norwood procedure by approximately 1.5:1. Table 2 summarizes the distribution of patient characteristics (demographics and preoperative and procedural factors) across groups with shunt failure and no shunt failure. Univariate analysis of patient characteristics by group showed no significant difference in median age at the time of the operation (7 days, p ¼ 0.404). Shunt failure was more prevalent among female patients than male patients (8.4% vs 6.6%, p ¼ 0.003; see Supplemental Results). The weight-for-age z-score and weight at the operation were both lower in the shunt failure group. Shunt failure was more prevalent among patients with noncardiac anatomic abnormalities or syndromes and among those with STS-CHSD preoperative factors, including preoperative mechanical ventilation. The prevalence of shunt failure did not differ between those
Early Shunt Failure Across Centers The number of eligible patients varied across centers with median 54 (interquartile range: 24, 114). The number of shunt failures by center was median 4 (interquartile range: 1, 8). The event rate for in-hospital shunt failure varied across centers with median 6.2% (interquartile range: 3.3%, 10.6%). The number of eligible shunt patients per center as well as failure rates varied considerably, without a clear association between the two (Supplemental Fig 1).
Risk of Early Shunt Failure Table 3 reports the results of the multivariable analysis. Risk factors for in-hospital shunt failure included lower weight at operation in neonates (OR, 1.35; p ¼ 0.0007) and infants (OR, 1.34; p ¼ 0.0002), preoperative hypercoagulable state (OR, 2.47; p ¼ 0.0306), and presence of any other STS-CHSD preoperative risk factors (OR, 1.24; p ¼ 0.0378). Shunt failure was more likely with a systemic artery-to-pulmonary artery shunt than with a systemic ventricle-to-pulmonary artery shunt. Neither use nor nonuse of cardiopulmonary bypass was a risk factor for shunt failure. Risk of shunt failure was not increased in single-ventricle patients, including those with HLHS.
Results of Sensitivity analysis Results of the multivariable analysis were unchanged after all patients with systemic ventricle-to-pulmonary artery shunts (Norwood, Sano modification) were excluded, with the same covariates being identified as risk factors for inhospital shunt failure. Results of the sensitivity analysis are reported, alongside results of the primary multivariable analysis, in Supplemental Table 3. The rate of occurrence of early shunt failure for the sensitivity analysis
Table 1. Shunt Type and Diagnostic Group: Occurrence and Rate of Shunt Failure
Variable Shunt type mBTS Central shunt Norwood Systemic artery to pulmonary Ventricle to pulmonary Unknown PBF source TAPVC þ shunt Other or unspecified shunt Diagnosis HLHS Non-HLHS single ventricle Other HLHS ¼ hypoplastic left heart syndrome; pulmonary venous connection.
Overall (N ¼ 9,172) (No.)
No Shunt Failure (n ¼ 8,498) (No.)
Shunt Failure (n ¼ 674) (No.)
Event Rate (Shunt Failure) (%)
3,474 1,202
3,217 1,094
257 108
7.4 9.0
1,466 2,233 232 118 447
1,328 2,117 218 105 419
138 116 14 13 28
9.4 5.2 6.0 11.0 6.3
2,984 2,100 4,088
2,788 1,937 3,773
196 163 315
6.6 7.8 7.7
mBTS ¼ modified Blalock-Taussig shunt;
PBF ¼ pulmonary blood flow;
TAPVC ¼ total anomalous
4
Variablea
Shunt Failure (n ¼ 674)
No Shunt Failure (n ¼ 8,498)
7.0 (5.0, 17.0)
7.0 (5.0, 17.0)
7.0 (5.0, 18.0)
7,405 1,767
560 114
6,845 1,653
7.6 6.5
0.108
5,377 3,785 4,849 –1.0 (–1.7, –0.4) 3.3 (2.9, 3.7)
355 318 362 –1.2 (–1.9, –0.5) 3.1 (2.8, 3.5)
5,022 3,467 4,487 –1.0 (–1.7, –0.4) 3.3 (2.9, 3.7)
6.6 8.4 7.5
0.003
6,556 2,616
477 197
6,079 2,419
7.3 7.5
0.673
8,358 469 206 1,378 463 1,249 1,875 4,142 60 217 173 2,528 53 2,475
615 36 14 104 48 101 159 339 8 16 15 222 8 206
7,743 433 192 1,274 415 1,148 1,716 3,803 52 201 158 2,306 45 2,269
7.4 7.7 6.8 7.5 10 8.1 8.5 8.2 13 7.4 8.7 8.8 15 8.3
0.921
a Continuous/ordinal data are presented as median (interquartile range) and categoric data as number of patients. based group means score statistics (equivalent to Wilcoxon tests) for all continuous/ordinal variables.
CT ¼ cardiothoracic;
Pre-op ¼ preoperative;
p Valueb 0.404
0.621 0.001 <.001
0.809 0.010 0.281 0.033 0.005 0.074 0.986 0.499 0.001 0.030 0.028
The p values are based on Pearson c2 tests for all categoric variables and on the c2 rank-
STS-CHSD ¼ The Society of Thoracic Surgeons Congenital Heart Surgery Database.
Ann Thorac Surg 2017;-:-–-
CPB ¼ cardiopulmonary bypass;
b
Event Rate (Shunt Failure) (%)
DO ET AL
Age at operation, d Age groups Neonates Infants Gender Male Female Non-Hispanic white Weight-for-age z-score Weight at operation, kg Operation type CPB No CPB Previous CT operations None 1 2 Prematurity Any noncardiac anatomic anomaly Any chromosomal abnormality Any syndrome Any STS-CHSD pre-op factors Pre-op mechanical circulatory support Shock, persistent at time of operation Pre-op renal dysfunction, renal failure requiring dialysis Pre-op mechanical ventilation Pre-op hypercoagulable state Any other pre-op factors
Overall (N ¼ 9,172)
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
Table 2. Patient Characteristics and Observed Occurrence of Early Shunt Failure (Unadjusted)
Ann Thorac Surg 2017;-:-–-
RICHARD E. CLARK AWARD DO ET AL EARLY SHUNT FAILURE
5
Table 3. Results of Multivariable Analysis of Associations With Early Shunt Failurea Adjustedb
Unadjusted Risk Factor
OR (95% CI)
Prematurity Previous cardiac operations Any noncardiac abnormalities Any genetic abnormalities Pre-op mechanical circulatory support Shock, persistent at time of operation Pre-op renal failure requiring dialysis Pre-op mechanical ventilatory support Pre-op neurologic deficit Pre-op coagulopathy Hypercoagulable state Intrinsic hypocoagulable state Any other pre-op risk factor Weight (per kg decrease) Neonates Infants Neonate vs infant Operation type (CPB vs non-CPB) Shunt type Central vs mBT Norwood–mBT vs mBT Norwood–RV-PA vs mBT Other vs mBT TAPVC þ shunt vs mBT Fundamental diagnosis HLHS vs other Non-HLHS single-ventricle vs other
1.03 0.89 1.57 1.25 1.99 1.01 1.21 1.29 1.23
p Value
p Value
(0.82–1.28) (0.65–1.23) (1.13–2.17) (1.04–1.50) (0.91–4.36) (0.56–1.83) (0.70–2.11) (1.08–1.54) (0.58–2.59)
0.8257 0.4741 0.0068 0.0185 0.0856 0.9713 0.4928 0.0059 0.5871
0.80 1.37 1.29 1.09 1.45 0.85 1.03 1.12 1.11
(0.62–1.04) (0.90–2.08) (0.91–1.82) (0.89–1.33) (0.62–3.36) (0.46–1.57) (0.58–1.82) (0.93–1.35) (0.51–2.39)
0.0909 0.1398 0.1569 0.3945 0.3878 0.6067 0.9207 0.2451 0.7986
2.71 (1.22–6.04) 1.02 (0.40–2.61) 1.32 (1.09–1.60)
0.0143 0.9737 0.0046
2.47 (1.09–5.60) 0.78 (0.29–2.07) 1.24 (1.01–1.53)
0.0306 0.6187 0.0378
1.00 1.06 1.21 0.95
(0.95–1.06) (1.02–1.11) (0.97–1.50) (0.79–1.15)
0.8885 0.0062 0.0884 0.6199
1.35 1.34 1.04 1.05
(1.13–1.61) (1.15–1.56) (0.46–2.34) (0.83–1.33)
0.0007 0.0002 0.9288 0.6857
1.30 1.27 0.66 0.81 1.57
(1.01–1.69) (1.00–1.61) (0.52–0.84) (0.54–1.23) (0.85–2.90)
0.0442 0.0471 0.0006 0.3286 0.1459
1.30 1.27 0.65 1.07 1.20
(0.99–1.69) (0.91–1.76) (0.45–0.93) (0.68–1.69) (0.63–2.30)
0.0569 0.1578 0.0200 0.7654 0.5754
0.81 (0.66–0.98) 1.03 (0.84–1.27)
0.0306 0.7815
0.97 (0.72–1.33) 1.02 (0.82–1.27)
0.8709 0.8490
a Cases of Norwood with source of pulmonary blood flow unspecified are excluded from models. association.
CI ¼ confidence interval; shunt; OR ¼ odds ratio; venous connection.
OR (95% CI)
b
Bold values indicate a statistically significant
CPB ¼ cardiopulmonary bypass; HLHS ¼ hypoplastic left heart syndrome; mBT ¼ modified Blalock-Taussig Pre-op ¼ preoperative; RV-PA ¼ right ventricle-to-pulmonary artery; TAPVC ¼ total anomalous pulmonary
group (right ventricle-to-pulmonary artery Norwood excluded) was 8.0% (558 of 6,939), compared with 7.3% (674 of 9,172) for the entire cohort. This difference is explained by the rate of early shunt failure of 5.2% in the systemic ventricle-to-pulmonary artery shunt group.
Outcomes Patients with in-hospital shunt failure had significantly higher rates of operative mortality (31.9% vs 11.1%, p < 0.001) and major morbidity (88.4% vs 29.4%, p < 0.001) and a longer median postoperative length of stay among survivors (45 vs 22 days, p < 0.001; Table 4). After all patients with systemic ventricle-to-pulmonary artery shunt were excluded from the analytic cohort, the comparison of secondary end points remained essentially unchanged.
Comment This is the largest reported analysis of factors associated with postoperative in-hospital shunt failure in neonates and
infants with congenital heart disease and the first multicenter study to evaluate preoperative risk factors and patient characteristics associated with shunt failure. We found that 7.3% of at-risk infants experienced early (in-hospital) shunt failure. Multivariable analysis showed shunt failure was more likely in patients with lower weight and in those with STS-CHSD preoperative risk factors. Systemic ventricle-topulmonary artery shunts were less prone to early failure but still had a 5.2% rate of occurrence. Our analysis confirms the substantial morbidity and increased rate of operative mortality associated with in-hospital shunt failure. Single-institution studies have reported early mortality rates after systemic-to-pulmonary shunt operations ranging from 4% to 14% [11–14]. Petrucci and colleagues [15] used the STS-CHSD to evaluate morbidity and death after the modified Blalock-Taussig shunt in neonates. Among 1,273 patients at 82 centers (2002 to 2009), the discharge mortality rate was 7.2%. Nearly one-third of the deaths occurred within the first 24 hours after the operation, and three-fourths occurred within the first 30 days. Shunt failure, per se, was not a specific study end point.
6
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
DO ET AL
Ann Thorac Surg 2017;-:-–-
Table 4. Outcomes: Shunt Failure and No Shunt Failure Groups (Unadjusted)Outcomea Overallb (N ¼ 9,172) Discharge mortality Operative mortality Morbidity composite Post-op length of stay (days) for hospital survivors a
1,122 1,159 3,071 23.0
Data are presented as number (%) or as median (interquartile range).
b
(12.2) (12.6) (33.5) (12.0, 42.0)
Shunt Failureb (n ¼ 674) 215 215 569 45.0
(31.9) (31.9) (84.4) (27.0, 84.0)
No Shunt Failureb (n ¼ 8,498) 907 944 2,502 22.0
(10.7) (11.1) (29.4) (12.0, 39.0)
p < 0.001 for all row comparisons.
Post-op ¼ postoperative.
In the Single Ventricle Reconstruction Trial, the highest number of deaths within the first 12 months occurred during the hospitalization related to the Norwood procedure. The most common category of cause of death was cardiovascular; this was true for both shunt types. But more granular information concerning shunt failure was not analyzed or reported [16]. Few recent studies have evaluated early shunt failure as a primary end point. Guzzetta and colleagues [8] reviewed 207 patients who had undergone modified Blalock-Taussig shunt procedures and reported that inhospital shunt occlusion occurred in 14 patients (6.8%). O’Connor and colleagues [17] reported a rate of shunt reintervention before discharge of 9.7% among 207 who underwent shunt operations. Shunt failure is costly in lives and utilization of resources but is incompletely understood. Shunt failure occurs across a heterogeneous spectrum of diagnoses and procedures. It may be related to technical (surgical) factors, hemodynamic factors, hematologic factors, or asyet-unstudied patient factors (genetic, disease-related susceptibility, etc). Diminished pulmonary blood flow resulting from shunt obstruction or occlusion may come about secondary to surgical technique, patient anatomy, or physiologic alterations. But diminished pulmonary blood flow is not a universal common pathway for scenarios that may require shunt reintervention. Excessive pulmonary blood flow can also necessitate a shuntrelated intervention. Our data clearly show that shunt failure significantly increases not only mortality but also hospital length of stay among survivors. We observed that median postoperative length of stay among survivors was more than twice as long in those who experienced shunt failure. The cost of increased days in the hospital, most which may be days in intensive care, represents not only a financial burden but has also been shown to be related to other important outcomes such as neurodevelopmental status. In addition, major complications were markedly more common in the shunt failure group. Prior multicenter studies have focused on the evaluation of pharmacologic strategies intended to reduce the risk of shunt failure caused by thrombosis. In a large, observational prospective study of infants with systemicto-pulmonary shunts (including Norwood/Sano), Li and colleagues [4] reported 1-year event rates of 26% for mortality and 12% for shunt thrombosis. Their analysis
suggested that after patients with early death were excluded, patients receiving acetylsalicylic acid had a lower risk of shunt thrombosis. The contribution of shunt failure to early death was not estimated. The multicenter, multinational prospective randomized Clopidogrel to Lower Arterial Thrombotic Risk in Neonates and Infants Trial (CLARINET) study investigated whether the addition of clopidogrel to conventional therapy reduced deaths from any cause and morbidity related to the shunt among infants with placement of a systemicto-pulmonary artery shunt [5]. The primary efficacy end point was a composite of death or heart transplantation, shunt thrombosis, or performance of a cardiac procedure attributable to an event considered to be thrombotic in nature before age 120 days. The rate of the composite primary end point did not differ significantly between the clopidogrel group (19.1%) and the placebo group (20.5%). Clopidogrel treatment did not have any significant benefit in any subgroup, including subgroups defined by shunt type (including systemic artery-to-pulmonary artery, systemic ventricle-to-pulmonary artery [Sano], and stented arterial duct). The CLARINET investigators acknowledged the challenges of organizing a clinical trial in a heterogeneous population of infants with rare diseases and concluded that, “despite an aggressive enrollment campaign spread over 134 sites in 31 countries, our study did not have adequate statistical power to test the equivalence of clopidogrel and placebo” [5]. These previous multicenter studies included systemic ventricle-to-pulmonary artery shunts in their evaluations of strategies to mitigate shunt thrombosis. The present study undertook a sensitivity analysis (exclusion of systemic ventricle-to-pulmonary artery shunts) to ensure that inclusion of these cases was not a reason for any errors in interpretation of the primary analysis.
Limitations We found that preoperative hypercoagulable state was associated with increased risk of shunt failure. We cannot rule out the possibility that some centers may test for coagulopathies after shunt thrombosis and then retrospectively code this as a preoperative risk factor. This would introduce a bias, because patients without shunt thrombosis might not be tested. Emani and associates [18, 19] undertook preoperative hypercoagulability panel testing in neonates undergoing cardiac operations
Ann Thorac Surg 2017;-:-–-
and found that some abnormal test results may help identify neonates at high risk for thrombus formation after cardiac operations. Detailed data concerning results of preoperative coagulation assays are not collected in STS-CHSD. Another potential limitation pertains to coding of interventional catheter procedures. This is one of multiple criteria that make up the algorithm for shunt failure identification. The possibility of imprecise coding always exits, and undercoding of catheter procedures in the surgical database cannot be ruled out. Although that may contribute to underestimation of the primary end point, the possibility that a few of the catheter interventions may not have been primarily directed at shunt-related problems must also be considered. The great majority of catheter interventions occurred in patients who also had post-shunt surgical interventions, suggesting that the effect of possible coding error on study inferences is minimal. It is possible that a patient could die in-hospital without ascertainment of shunt failure as contributory to the death. Scenarios of death in the absence of investigation by echocardiography or other means (including mediastinal exploration for assessment of shunt status) and without extracorporeal life support and catheter investigation, are probably quite rare. We acknowledge that a few such cases would lead to slight underestimation of the rate of in-hospital shunt failure. It would, however, likewise lead to underestimation of the rate of mortality associated with shunt failure, lending further support to one of our conclusions; namely, that early shunt failure is an important problem with a high rate of associated death. Additional limitations related to the retrospective design and the voluntary nature of the database might affect the complete capture of all relevant data. Organization of the STS-CHSD does not lend itself to ascertainment of the time between occurrence of early shunt failure and outcomes/reinterventions. Also, we had no way to evaluate the potential influence of suboptimal surgical technique or the effect of medical care regimens. Some potentially important predictor variables that are not captured in the STS-CHSD include shunt size (diameter and length), protamine administration, and postoperative antiaggregant or antithrombotic prophylaxis regimens.
Conclusions The present study is the largest to date looking at shunt failure in infants, a complex and incompletely understood problem. Relying on a representative and heterogeneous population in a very inclusive clinical registry, this analysis illustrates the scope and magnitude of the problem and has further clarified the characteristics of patients who may be at greatest risk. On a day-to-day basis, this information may help to identify individual patients that warrant expectant surveillance, enhanced pharmacologic management, or other strategies to reduce the risk of shunt failure. Perhaps more importantly, it provides key
RICHARD E. CLARK AWARD DO ET AL EARLY SHUNT FAILURE
7
information that may be helpful in the design and development of future clinical trials or collaborative quality improvement initiatives designed to reduce the cost in lives and resources that is associated with early shunt dysfunction. Audio Discussion: Audio of the discussion that followed the presentation of this paper at the STS Annual Meeting can be accessed in the online version of this article [http://dx.doi.org/10.1016/j.athoracsur.2017. 06.028] on http://www.annalsthoracicsurgery.org.
References 1. Blalock A, Taussig HB. The surgical treatment of malformations of the heart in which there is pulmonary stenosis or pulmonary atresia. JAMA 1945;128:189–202. 2. Sano S, Ishino K, Kado H, et al. Outcome of right ventricleto-pulmonary artery shunt in first-stage palliation of hypoplastic left heart syndrome: a multi-institutional study. Ann Thorac Surg 2004;78:1951–7. 3. Pizarro C, Mroczek T, Malec E, Norwood WI. Right ventricle to pulmonary artery conduit reduces interim mortality after stage 1 Norwood for hypoplastic left heart syndrome. Ann Thorac Surg 2004;78:1959–63. 4. Li JS, Yow E, Berezny KY, et al. Clinical outcomes of palliative surgery including a systemic-to-pulmonary artery shunt in infants with cyanotic congenital heart disease: does aspirin make a difference? Circulation 2007;116: 293–7. 5. Wessel DL, Berger F, Li JS, et al, CLARINET Investigators. Clopidogrel in infants with systemic-to-pulmonary-artery shunts. N Engl J Med 2013;368:2377–84. 6. Heidari-Bateni G, Norouzi S, Hall M, Brar A, Eghtesady P. Defining the best practice patterns for the neonatal systemicto-pulmonary artery shunt procedure. J Thorac Cardiovasc Surg 2014;147:869–73. 7. Gedicke M, Morgan G, Parry A, Martin R, Tulloh R. Risk factors for acute shunt blockage in children after modified Blalock-Taussig shunt operations. Heart Vessels 2010;25: 405–9. 8. Guzzetta NA, Foster GS, Mruthinti N, Kilgore PD, Miller BE, Kanter KR. In-hospital shunt occlusion in infants undergoing a modified Blalock-Taussig shunt. Ann Thorac Surg 2013;96: 176–82. 9. Jacobs JP, Mayer JE Jr, Mavroudis C, et al. The Society of Thoracic Surgeons Congenital Heart Surgery Database: 2016 update on outcomes and quality. Ann Thorac Surg 2016;101: 850–62. 10. The Society of Thoracic Surgeons. STS Congenital Heart Surgery Database Data Specifications. Version 3.2. Available at: http://www.sts.org/sites/default/files/documents/ CongenitalDataSpecsV3_22.pdf. Accessed January 9, 2017. 11. Alkhulaifi AM, Lacour-Gayet F, Serraf A, Belli E, Planche C. Systemic pulmonary shunts in neonates: early clinical outcome and choice of surgical approach. Ann Thorac Surg 2000;69:1499–504. 12. Myers JW, Ghanayem NS, Cao Y, et al. Outcomes of systemic to pulmonary artery shunts in patients weighing less than 3 kg: analysis of shunt type, size, and surgical approach. J Thorac Cardiovasc Surg 2014;147:672–7. 13. Williams JA, Bansal AK, Kim BJ, et al. Two thousand BlalockTaussig shunts: a six-decade experience. Ann Thorac Surg 2007;84:2070–5. 14. Fenton KN, Siewers RD, Rebovich B, Pigula FA. Interim mortality in infants with systemic-to-pulmonary artery shunts. Ann Thorac Surg 2003;76:152–6.
8
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
DO ET AL
15. Petrucci O, O’Brien SM, Jacobs ML, Jacobs JP, Manning PB, Eghtesady P. Risk factors for mortality and morbidity after the neonatal Blalock-Taussig shunt procedure. Ann Thorac Surg 2011;92:642–51. 16. Ohye RG, Schonbeck JV, Eghtesady P, et al. Cause, timing, and location of death in the Single Ventricle Reconstruction trial. J Thorac Cardiovasc Surg 2012;144:907–14. 17. O’Connor MJ, Ravishankar C, Ballweg JA, et al. Early systemic-to-pulmonary artery shunt intervention in
Ann Thorac Surg 2017;-:-–-
neonates with congenital heart disease. J Thorac Cardiovasc Surg 2011;142:106–12. 18. Emani S, Zurakowski D, Baird CW, et al. Hypercoagulability panel testing predicts thrombosis in neonates undergoing cardiac surgery. Am J Hematol 2014;89:151–5. 19. Emani S, Trainor B, Zurakowski D, et al. Aspirin unresponsiveness predicts thrombosis in high-risk pediatric patients after cardiac surgery. J Thorac Cardiovasc Surg 2014;148: 810–4; discussion 814–6.
Shunt Failure—Risk Factors and Outcomes: An Analysis of The Society of Thoracic Surgeons Congenital Heart Surgery Database Nhue Do, MD, Kevin D. Hill, MD, MS, Amelia S. Wallace, MS, Luca Vricella, MD, Duke Cameron, MD, James Quintessenza, MD, Neil Goldenberg, MD, PhD, Constantine Mavroudis, MD, Tom Karl, MD, Sara K. Pasquali, MD, MS, Jeffrey P. Jacobs, MD, and Marshall L. Jacobs, MD Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania; Duke Clinical Research Institute, Durham, North Carolina; Division of Cardiac Surgery, Massachusetts General Hospital, Boston, Massachusetts; Cincinnati Children’s Hospital, Cincinnati, Ohio; Johns Hopkins All Children’s Hospital, St. Petersburg, Florida; Florida Hospital, Orlando, Florida; and C.S. Mott Children’s Hospital, Ann Arbor, Michigan
Background. Systemic-to-pulmonary shunt failure is a potentially catastrophic complication. We analyzed a large multicenter clinical registry to describe the prevalence and evaluate risk factors. Methods. Infants (aged £365 days) undergoing shunt operations (systemic artery-to-pulmonary artery or systemic ventricle-to-pulmonary artery) in The Society of Thoracic Surgeons Congenital Heart Surgery Database (STS-CHSD) from 2010 to 2015 were included. Multivariable logistic regression was used to evaluate risk factors for in-hospital shunt failure. Model covariates included patient characteristics, preoperative factors, procedural factors including shunt type, and center effects. Centers with more than 15% missing data for key covariates were excluded. Results. Shunt operations were performed in 9,172 infants (118 centers). In-hospital shunt failure occurred in 674 (7.3%). In multivariable analysis, risk factors for in-hospital shunt failure included lower weight at operation (odds ratio [OR], 1.35; p [ 0.001), preoperative hypercoagulable state (OR, 2.47; p [ 0.031), and the
presence of any other STS-CHSD preoperative risk factors (OR, 1.24; p [ 0.038). Shunt failure was less likely with a systemic ventricle-to-pulmonary artery shunt than a systemic artery-to-pulmonary artery shunt (OR, 0.65; p [ 0.020). Neither cardiopulmonary bypass nor singleventricle diagnosis was a risk factor for shunt failure. Patients with in-hospital shunt failure had significantly higher rates of operative mortality (31.9% vs 11.1%, p < 0.001) and major morbidity (84.4% vs 29.4%, p < 0.001), and longer median postoperative length of stay among survivors (45 vs 22 days, p < 0.001). Conclusions. In-hospital shunt failure is common, and associated mortality risk is high. These data highlight at-risk patients and procedural cohorts that warrant expectant surveillance and may benefit from enhanced antithrombotic prophylaxis or other management strategies to reduce shunt failure. These findings may inform planning of future clinical trials.
S
these patients, creation of a systemic ventricle-topulmonary artery shunt is an alternative to the traditional systemic artery-to-pulmonary artery shunt [2, 3]. Early failure of both categories of palliative systemic-topulmonary shunts is an incompletely understood phenomenon that accounts for important morbidity and death among infants and neonates. Multicenter studies and clinical trials have focused primarily on pharmacologic strategies to reduce the risk of shunt failure caused by thrombosis [4–6]. Their utility for
hunt operations connecting the pulmonary arteries to a systemic artery or the systemic ventricle are common in the management of neonates and infants with congenital heart disease. Originally introduced to treat patients with cyanosis related to diminished pulmonary blood flow [1], systemic-to-pulmonary artery shunts are also used to regulate pulmonary blood flow, especially for patients undergoing palliation of single-ventricle anomalies such as hypoplastic left heart syndrome (HLHS). For
(Ann Thorac Surg 2017;-:-–-) Ó 2017 by The Society of Thoracic Surgeons
Accepted for publication June 6, 2017. Presented at the Fifty-third Annual Meeting of the Society of Thoracic Surgeons, Houston, TX, Jan 21–25, 2017. Winner of the Richard E. Clark Award for Congenital Heart Surgery. Address correspondence to Dr Jacobs, Division of Cardiac Surgery, The Johns Hopkins Hospital, Zayed 7107, 1800 Orleans St, Baltimore, MD 21287; email: [email protected].
Ó 2017 by The Society of Thoracic Surgeons Published by Elsevier Inc.
The Supplemental Tables and Figure can be viewed in the online version of this article [http://dx.doi.org/10. 1016/j.athoracsur.2017.06.028] on http://www.annals thoracicsurgery.org.
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2017.06.028
2
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
DO ET AL
guiding clinical decision making has been limited. Some have been underpowered, and some have had limited risk adjustment of patients. Many single-center studies have not considered the multifactorial nature of early shunt failure, focusing primarily on thrombosis [7, 8]. Using data from The Society of Thoracic Surgeons Congenital Heart Surgery Database (STS-CHSD), we sought to characterize the group of patients who were observed to have experienced early in-hospital shunt failure. This information should be helpful in risk assessment of individual patients and may enhance the precision and productivity of future prospective studies by informing eligibility criteria, risk stratification, and covariate adjustment.
Patients and Methods Data Source The STS-CHSD collects data on neonates, infants, children, and adults undergoing operations for congenital heart disease, currently including more than 400,000 operations conducted at 127 centers in North America. The database encompasses more than 95% of United States centers performing pediatric heart operations and approximately 98% of all pediatric cardiac operations done annually in the United States [9]. Coding is accomplished by clinicians and ancillary support staff using International Pediatric and Congenital Cardiac Code terminology, including demographics, anatomic diagnoses, noncardiac abnormalities, preoperative risk factors, intraoperative details, surgical procedures, postoperative complications, postoperative length of stay, and operative mortality. Data on all patients undergoing congenital heart operations at participating centers are entered into STS-CHSD data collection software. The Duke Clinical Research Institute is the data warehouse and analysis center for STS national databases. Evaluation of data quality includes intrinsic data verification and a formal process of site visits and data audits conducted by independent data quality personnel and pediatric cardiac surgeons at 10% of participating institutions annually. The STS-CHS Database Access and Publications Committee and the Duke University Institutional Review Board approved the study. The study was not considered human subjects research by the Duke University Institutional Review Board in accordance with the Common Rule (45 CFR 46.102[f]).
Ann Thorac Surg 2017;-:-–-
centers with more than 15% missing data for key covariates were excluded. Patients were considered to have experienced early shunt failure if one or more of three criteria were met: (1) diagnosis of “shunt failure” or (2) procedure code “shunt, reoperation” or (3) any “additional procedure” from a list of specific shunt interventions, entered in association with an event subsequent to the eligible index operation (Supplemental Table 2).
End Points The primary end point was ascertainment of early shunt failure, as defined above. Secondary end points were operative mortality, discharge mortality, and composite morbidity. Composite morbidity was defined as occurrence of any one or more of the following complications: renal failure requiring temporary or permanent dialysis, neurologic deficit persisting at discharge, arrhythmia necessitating permanent pacemaker, postoperative mechanical circulatory support, paralyzed diaphragm, and cardiac arrest. Definitions of the individual variables captured in the composite end point can be found in the STS CHSD specifications at the STS Web site [10].
Analysis Characteristics of patients, including distributions of demographic, preoperative, and operative factors, were explored and described using standard summary statistics. Pearson c2 tests for categoric variables and c2 rank-based group mean score statistics (equivalent to Wilcoxon tests) for continuous/ordinal variables were used to assess univariate differences in the distribution of patient characteristics across groups with early shunt failure and no shunt failure. Multivariable analysis was used to evaluate associations between potential risk factors and the occurrence of early shunt failure (as defined above). Conditional logistic regression stratified by center was used to account for center-level variability in practices. Potential risk factors for shunt failure included in the models as covariates were defined a priori based on the clinical judgement of the investigative team. Model covariates included patient characteristics in the categories listed above, in addition to the mutually exclusive shunt procedure types and diagnostic categories (HLHS, non-HLHS single ventricle, other). Results are summarized using estimated odds ratios (ORs), confidence intervals of the estimated ORs, and p values.
Patient Population
Sensitivity Analysis
Neonates and infants (aged 0 to 365 days) undergoing cardiac operations from January 1, 2010, through December 31, 2015, at STS-CHSD centers were potentially eligible for this study. The study population consisted of all patients for whom the index operation included one or more of the systemic-to-pulmonary shunt procedures included in STS CHSD data collection versions 3.0 or 3.22, or both (Supplemental Table 1). Hybrid Norwood procedures and combined Norwood and Rastelli (Yasui) procedures were excluded. Data from
Analyses were repeated after exclusion of all patients whose shunt type was a systemic ventricle-to-pulmonary artery shunt (Norwood with “Sano modification”; ie, ventricle-to-pulmonary artery conduit). This enabled us to estimate the prevalence of early shunt failure and assess associations of patient and procedural factors and outcome variables in a cohort limited to systemic arteryto-pulmonary artery shunt operations. Analyses were performed using SAS software (SAS Institute Inc, Cary, NC).
Ann Thorac Surg 2017;-:-–-
RICHARD E. CLARK AWARD DO ET AL EARLY SHUNT FAILURE
Results
3
with or without cardiopulmonary bypass support for the shunt procedure.
Study Population Characteristics A total of 9,172 patients from 118 institutions were eligible for analysis and considered to be potentially at risk for early shunt failure. In-hospital shunt failure was identified in 674 patients (7.3%) of the 9,172 who had undergone a shunt procedure as part of the index operation. Of the three study criteria for ascertainment of early shunt failure (Supplemental Table 2), diagnosis of “shunt failure” was coded in 267 patients; procedure “shunt, reoperation” was coded in 328, and the “additional procedure” criteria were met in 398. In 273 instances, the patient met two or all three of the criteria. The types of shunts performed and the occurrence of shunt failure for each type are reported in Table 1. The most commonly performed was the modified BlalockTaussig shunt (mBTS) in a setting other than a Norwood procedure, with 3,474 (37.9%) overall. The second most common was a systemic ventricle-to-pulmonary artery shunt (Sano) as part of a Norwood procedure. These outnumbered systemic artery-to-pulmonary artery shunts as part of a Norwood procedure by approximately 1.5:1. Table 2 summarizes the distribution of patient characteristics (demographics and preoperative and procedural factors) across groups with shunt failure and no shunt failure. Univariate analysis of patient characteristics by group showed no significant difference in median age at the time of the operation (7 days, p ¼ 0.404). Shunt failure was more prevalent among female patients than male patients (8.4% vs 6.6%, p ¼ 0.003; see Supplemental Results). The weight-for-age z-score and weight at the operation were both lower in the shunt failure group. Shunt failure was more prevalent among patients with noncardiac anatomic abnormalities or syndromes and among those with STS-CHSD preoperative factors, including preoperative mechanical ventilation. The prevalence of shunt failure did not differ between those
Early Shunt Failure Across Centers The number of eligible patients varied across centers with median 54 (interquartile range: 24, 114). The number of shunt failures by center was median 4 (interquartile range: 1, 8). The event rate for in-hospital shunt failure varied across centers with median 6.2% (interquartile range: 3.3%, 10.6%). The number of eligible shunt patients per center as well as failure rates varied considerably, without a clear association between the two (Supplemental Fig 1).
Risk of Early Shunt Failure Table 3 reports the results of the multivariable analysis. Risk factors for in-hospital shunt failure included lower weight at operation in neonates (OR, 1.35; p ¼ 0.0007) and infants (OR, 1.34; p ¼ 0.0002), preoperative hypercoagulable state (OR, 2.47; p ¼ 0.0306), and presence of any other STS-CHSD preoperative risk factors (OR, 1.24; p ¼ 0.0378). Shunt failure was more likely with a systemic artery-to-pulmonary artery shunt than with a systemic ventricle-to-pulmonary artery shunt. Neither use nor nonuse of cardiopulmonary bypass was a risk factor for shunt failure. Risk of shunt failure was not increased in single-ventricle patients, including those with HLHS.
Results of Sensitivity analysis Results of the multivariable analysis were unchanged after all patients with systemic ventricle-to-pulmonary artery shunts (Norwood, Sano modification) were excluded, with the same covariates being identified as risk factors for inhospital shunt failure. Results of the sensitivity analysis are reported, alongside results of the primary multivariable analysis, in Supplemental Table 3. The rate of occurrence of early shunt failure for the sensitivity analysis
Table 1. Shunt Type and Diagnostic Group: Occurrence and Rate of Shunt Failure
Variable Shunt type mBTS Central shunt Norwood Systemic artery to pulmonary Ventricle to pulmonary Unknown PBF source TAPVC þ shunt Other or unspecified shunt Diagnosis HLHS Non-HLHS single ventricle Other HLHS ¼ hypoplastic left heart syndrome; pulmonary venous connection.
Overall (N ¼ 9,172) (No.)
No Shunt Failure (n ¼ 8,498) (No.)
Shunt Failure (n ¼ 674) (No.)
Event Rate (Shunt Failure) (%)
3,474 1,202
3,217 1,094
257 108
7.4 9.0
1,466 2,233 232 118 447
1,328 2,117 218 105 419
138 116 14 13 28
9.4 5.2 6.0 11.0 6.3
2,984 2,100 4,088
2,788 1,937 3,773
196 163 315
6.6 7.8 7.7
mBTS ¼ modified Blalock-Taussig shunt;
PBF ¼ pulmonary blood flow;
TAPVC ¼ total anomalous
4
Variablea
Shunt Failure (n ¼ 674)
No Shunt Failure (n ¼ 8,498)
7.0 (5.0, 17.0)
7.0 (5.0, 17.0)
7.0 (5.0, 18.0)
7,405 1,767
560 114
6,845 1,653
7.6 6.5
0.108
5,377 3,785 4,849 –1.0 (–1.7, –0.4) 3.3 (2.9, 3.7)
355 318 362 –1.2 (–1.9, –0.5) 3.1 (2.8, 3.5)
5,022 3,467 4,487 –1.0 (–1.7, –0.4) 3.3 (2.9, 3.7)
6.6 8.4 7.5
0.003
6,556 2,616
477 197
6,079 2,419
7.3 7.5
0.673
8,358 469 206 1,378 463 1,249 1,875 4,142 60 217 173 2,528 53 2,475
615 36 14 104 48 101 159 339 8 16 15 222 8 206
7,743 433 192 1,274 415 1,148 1,716 3,803 52 201 158 2,306 45 2,269
7.4 7.7 6.8 7.5 10 8.1 8.5 8.2 13 7.4 8.7 8.8 15 8.3
0.921
a Continuous/ordinal data are presented as median (interquartile range) and categoric data as number of patients. based group means score statistics (equivalent to Wilcoxon tests) for all continuous/ordinal variables.
CT ¼ cardiothoracic;
Pre-op ¼ preoperative;
p Valueb 0.404
0.621 0.001 <.001
0.809 0.010 0.281 0.033 0.005 0.074 0.986 0.499 0.001 0.030 0.028
The p values are based on Pearson c2 tests for all categoric variables and on the c2 rank-
STS-CHSD ¼ The Society of Thoracic Surgeons Congenital Heart Surgery Database.
Ann Thorac Surg 2017;-:-–-
CPB ¼ cardiopulmonary bypass;
b
Event Rate (Shunt Failure) (%)
DO ET AL
Age at operation, d Age groups Neonates Infants Gender Male Female Non-Hispanic white Weight-for-age z-score Weight at operation, kg Operation type CPB No CPB Previous CT operations None 1 2 Prematurity Any noncardiac anatomic anomaly Any chromosomal abnormality Any syndrome Any STS-CHSD pre-op factors Pre-op mechanical circulatory support Shock, persistent at time of operation Pre-op renal dysfunction, renal failure requiring dialysis Pre-op mechanical ventilation Pre-op hypercoagulable state Any other pre-op factors
Overall (N ¼ 9,172)
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
Table 2. Patient Characteristics and Observed Occurrence of Early Shunt Failure (Unadjusted)
Ann Thorac Surg 2017;-:-–-
RICHARD E. CLARK AWARD DO ET AL EARLY SHUNT FAILURE
5
Table 3. Results of Multivariable Analysis of Associations With Early Shunt Failurea Adjustedb
Unadjusted Risk Factor
OR (95% CI)
Prematurity Previous cardiac operations Any noncardiac abnormalities Any genetic abnormalities Pre-op mechanical circulatory support Shock, persistent at time of operation Pre-op renal failure requiring dialysis Pre-op mechanical ventilatory support Pre-op neurologic deficit Pre-op coagulopathy Hypercoagulable state Intrinsic hypocoagulable state Any other pre-op risk factor Weight (per kg decrease) Neonates Infants Neonate vs infant Operation type (CPB vs non-CPB) Shunt type Central vs mBT Norwood–mBT vs mBT Norwood–RV-PA vs mBT Other vs mBT TAPVC þ shunt vs mBT Fundamental diagnosis HLHS vs other Non-HLHS single-ventricle vs other
1.03 0.89 1.57 1.25 1.99 1.01 1.21 1.29 1.23
p Value
p Value
(0.82–1.28) (0.65–1.23) (1.13–2.17) (1.04–1.50) (0.91–4.36) (0.56–1.83) (0.70–2.11) (1.08–1.54) (0.58–2.59)
0.8257 0.4741 0.0068 0.0185 0.0856 0.9713 0.4928 0.0059 0.5871
0.80 1.37 1.29 1.09 1.45 0.85 1.03 1.12 1.11
(0.62–1.04) (0.90–2.08) (0.91–1.82) (0.89–1.33) (0.62–3.36) (0.46–1.57) (0.58–1.82) (0.93–1.35) (0.51–2.39)
0.0909 0.1398 0.1569 0.3945 0.3878 0.6067 0.9207 0.2451 0.7986
2.71 (1.22–6.04) 1.02 (0.40–2.61) 1.32 (1.09–1.60)
0.0143 0.9737 0.0046
2.47 (1.09–5.60) 0.78 (0.29–2.07) 1.24 (1.01–1.53)
0.0306 0.6187 0.0378
1.00 1.06 1.21 0.95
(0.95–1.06) (1.02–1.11) (0.97–1.50) (0.79–1.15)
0.8885 0.0062 0.0884 0.6199
1.35 1.34 1.04 1.05
(1.13–1.61) (1.15–1.56) (0.46–2.34) (0.83–1.33)
0.0007 0.0002 0.9288 0.6857
1.30 1.27 0.66 0.81 1.57
(1.01–1.69) (1.00–1.61) (0.52–0.84) (0.54–1.23) (0.85–2.90)
0.0442 0.0471 0.0006 0.3286 0.1459
1.30 1.27 0.65 1.07 1.20
(0.99–1.69) (0.91–1.76) (0.45–0.93) (0.68–1.69) (0.63–2.30)
0.0569 0.1578 0.0200 0.7654 0.5754
0.81 (0.66–0.98) 1.03 (0.84–1.27)
0.0306 0.7815
0.97 (0.72–1.33) 1.02 (0.82–1.27)
0.8709 0.8490
a Cases of Norwood with source of pulmonary blood flow unspecified are excluded from models. association.
CI ¼ confidence interval; shunt; OR ¼ odds ratio; venous connection.
OR (95% CI)
b
Bold values indicate a statistically significant
CPB ¼ cardiopulmonary bypass; HLHS ¼ hypoplastic left heart syndrome; mBT ¼ modified Blalock-Taussig Pre-op ¼ preoperative; RV-PA ¼ right ventricle-to-pulmonary artery; TAPVC ¼ total anomalous pulmonary
group (right ventricle-to-pulmonary artery Norwood excluded) was 8.0% (558 of 6,939), compared with 7.3% (674 of 9,172) for the entire cohort. This difference is explained by the rate of early shunt failure of 5.2% in the systemic ventricle-to-pulmonary artery shunt group.
Outcomes Patients with in-hospital shunt failure had significantly higher rates of operative mortality (31.9% vs 11.1%, p < 0.001) and major morbidity (88.4% vs 29.4%, p < 0.001) and a longer median postoperative length of stay among survivors (45 vs 22 days, p < 0.001; Table 4). After all patients with systemic ventricle-to-pulmonary artery shunt were excluded from the analytic cohort, the comparison of secondary end points remained essentially unchanged.
Comment This is the largest reported analysis of factors associated with postoperative in-hospital shunt failure in neonates and
infants with congenital heart disease and the first multicenter study to evaluate preoperative risk factors and patient characteristics associated with shunt failure. We found that 7.3% of at-risk infants experienced early (in-hospital) shunt failure. Multivariable analysis showed shunt failure was more likely in patients with lower weight and in those with STS-CHSD preoperative risk factors. Systemic ventricle-topulmonary artery shunts were less prone to early failure but still had a 5.2% rate of occurrence. Our analysis confirms the substantial morbidity and increased rate of operative mortality associated with in-hospital shunt failure. Single-institution studies have reported early mortality rates after systemic-to-pulmonary shunt operations ranging from 4% to 14% [11–14]. Petrucci and colleagues [15] used the STS-CHSD to evaluate morbidity and death after the modified Blalock-Taussig shunt in neonates. Among 1,273 patients at 82 centers (2002 to 2009), the discharge mortality rate was 7.2%. Nearly one-third of the deaths occurred within the first 24 hours after the operation, and three-fourths occurred within the first 30 days. Shunt failure, per se, was not a specific study end point.
6
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
DO ET AL
Ann Thorac Surg 2017;-:-–-
Table 4. Outcomes: Shunt Failure and No Shunt Failure Groups (Unadjusted)Outcomea Overallb (N ¼ 9,172) Discharge mortality Operative mortality Morbidity composite Post-op length of stay (days) for hospital survivors a
1,122 1,159 3,071 23.0
Data are presented as number (%) or as median (interquartile range).
b
(12.2) (12.6) (33.5) (12.0, 42.0)
Shunt Failureb (n ¼ 674) 215 215 569 45.0
(31.9) (31.9) (84.4) (27.0, 84.0)
No Shunt Failureb (n ¼ 8,498) 907 944 2,502 22.0
(10.7) (11.1) (29.4) (12.0, 39.0)
p < 0.001 for all row comparisons.
Post-op ¼ postoperative.
In the Single Ventricle Reconstruction Trial, the highest number of deaths within the first 12 months occurred during the hospitalization related to the Norwood procedure. The most common category of cause of death was cardiovascular; this was true for both shunt types. But more granular information concerning shunt failure was not analyzed or reported [16]. Few recent studies have evaluated early shunt failure as a primary end point. Guzzetta and colleagues [8] reviewed 207 patients who had undergone modified Blalock-Taussig shunt procedures and reported that inhospital shunt occlusion occurred in 14 patients (6.8%). O’Connor and colleagues [17] reported a rate of shunt reintervention before discharge of 9.7% among 207 who underwent shunt operations. Shunt failure is costly in lives and utilization of resources but is incompletely understood. Shunt failure occurs across a heterogeneous spectrum of diagnoses and procedures. It may be related to technical (surgical) factors, hemodynamic factors, hematologic factors, or asyet-unstudied patient factors (genetic, disease-related susceptibility, etc). Diminished pulmonary blood flow resulting from shunt obstruction or occlusion may come about secondary to surgical technique, patient anatomy, or physiologic alterations. But diminished pulmonary blood flow is not a universal common pathway for scenarios that may require shunt reintervention. Excessive pulmonary blood flow can also necessitate a shuntrelated intervention. Our data clearly show that shunt failure significantly increases not only mortality but also hospital length of stay among survivors. We observed that median postoperative length of stay among survivors was more than twice as long in those who experienced shunt failure. The cost of increased days in the hospital, most which may be days in intensive care, represents not only a financial burden but has also been shown to be related to other important outcomes such as neurodevelopmental status. In addition, major complications were markedly more common in the shunt failure group. Prior multicenter studies have focused on the evaluation of pharmacologic strategies intended to reduce the risk of shunt failure caused by thrombosis. In a large, observational prospective study of infants with systemicto-pulmonary shunts (including Norwood/Sano), Li and colleagues [4] reported 1-year event rates of 26% for mortality and 12% for shunt thrombosis. Their analysis
suggested that after patients with early death were excluded, patients receiving acetylsalicylic acid had a lower risk of shunt thrombosis. The contribution of shunt failure to early death was not estimated. The multicenter, multinational prospective randomized Clopidogrel to Lower Arterial Thrombotic Risk in Neonates and Infants Trial (CLARINET) study investigated whether the addition of clopidogrel to conventional therapy reduced deaths from any cause and morbidity related to the shunt among infants with placement of a systemicto-pulmonary artery shunt [5]. The primary efficacy end point was a composite of death or heart transplantation, shunt thrombosis, or performance of a cardiac procedure attributable to an event considered to be thrombotic in nature before age 120 days. The rate of the composite primary end point did not differ significantly between the clopidogrel group (19.1%) and the placebo group (20.5%). Clopidogrel treatment did not have any significant benefit in any subgroup, including subgroups defined by shunt type (including systemic artery-to-pulmonary artery, systemic ventricle-to-pulmonary artery [Sano], and stented arterial duct). The CLARINET investigators acknowledged the challenges of organizing a clinical trial in a heterogeneous population of infants with rare diseases and concluded that, “despite an aggressive enrollment campaign spread over 134 sites in 31 countries, our study did not have adequate statistical power to test the equivalence of clopidogrel and placebo” [5]. These previous multicenter studies included systemic ventricle-to-pulmonary artery shunts in their evaluations of strategies to mitigate shunt thrombosis. The present study undertook a sensitivity analysis (exclusion of systemic ventricle-to-pulmonary artery shunts) to ensure that inclusion of these cases was not a reason for any errors in interpretation of the primary analysis.
Limitations We found that preoperative hypercoagulable state was associated with increased risk of shunt failure. We cannot rule out the possibility that some centers may test for coagulopathies after shunt thrombosis and then retrospectively code this as a preoperative risk factor. This would introduce a bias, because patients without shunt thrombosis might not be tested. Emani and associates [18, 19] undertook preoperative hypercoagulability panel testing in neonates undergoing cardiac operations
Ann Thorac Surg 2017;-:-–-
and found that some abnormal test results may help identify neonates at high risk for thrombus formation after cardiac operations. Detailed data concerning results of preoperative coagulation assays are not collected in STS-CHSD. Another potential limitation pertains to coding of interventional catheter procedures. This is one of multiple criteria that make up the algorithm for shunt failure identification. The possibility of imprecise coding always exits, and undercoding of catheter procedures in the surgical database cannot be ruled out. Although that may contribute to underestimation of the primary end point, the possibility that a few of the catheter interventions may not have been primarily directed at shunt-related problems must also be considered. The great majority of catheter interventions occurred in patients who also had post-shunt surgical interventions, suggesting that the effect of possible coding error on study inferences is minimal. It is possible that a patient could die in-hospital without ascertainment of shunt failure as contributory to the death. Scenarios of death in the absence of investigation by echocardiography or other means (including mediastinal exploration for assessment of shunt status) and without extracorporeal life support and catheter investigation, are probably quite rare. We acknowledge that a few such cases would lead to slight underestimation of the rate of in-hospital shunt failure. It would, however, likewise lead to underestimation of the rate of mortality associated with shunt failure, lending further support to one of our conclusions; namely, that early shunt failure is an important problem with a high rate of associated death. Additional limitations related to the retrospective design and the voluntary nature of the database might affect the complete capture of all relevant data. Organization of the STS-CHSD does not lend itself to ascertainment of the time between occurrence of early shunt failure and outcomes/reinterventions. Also, we had no way to evaluate the potential influence of suboptimal surgical technique or the effect of medical care regimens. Some potentially important predictor variables that are not captured in the STS-CHSD include shunt size (diameter and length), protamine administration, and postoperative antiaggregant or antithrombotic prophylaxis regimens.
Conclusions The present study is the largest to date looking at shunt failure in infants, a complex and incompletely understood problem. Relying on a representative and heterogeneous population in a very inclusive clinical registry, this analysis illustrates the scope and magnitude of the problem and has further clarified the characteristics of patients who may be at greatest risk. On a day-to-day basis, this information may help to identify individual patients that warrant expectant surveillance, enhanced pharmacologic management, or other strategies to reduce the risk of shunt failure. Perhaps more importantly, it provides key
RICHARD E. CLARK AWARD DO ET AL EARLY SHUNT FAILURE
7
information that may be helpful in the design and development of future clinical trials or collaborative quality improvement initiatives designed to reduce the cost in lives and resources that is associated with early shunt dysfunction. Audio Discussion: Audio of the discussion that followed the presentation of this paper at the STS Annual Meeting can be accessed in the online version of this article [http://dx.doi.org/10.1016/j.athoracsur.2017. 06.028] on http://www.annalsthoracicsurgery.org.
References 1. Blalock A, Taussig HB. The surgical treatment of malformations of the heart in which there is pulmonary stenosis or pulmonary atresia. JAMA 1945;128:189–202. 2. Sano S, Ishino K, Kado H, et al. Outcome of right ventricleto-pulmonary artery shunt in first-stage palliation of hypoplastic left heart syndrome: a multi-institutional study. Ann Thorac Surg 2004;78:1951–7. 3. Pizarro C, Mroczek T, Malec E, Norwood WI. Right ventricle to pulmonary artery conduit reduces interim mortality after stage 1 Norwood for hypoplastic left heart syndrome. Ann Thorac Surg 2004;78:1959–63. 4. Li JS, Yow E, Berezny KY, et al. Clinical outcomes of palliative surgery including a systemic-to-pulmonary artery shunt in infants with cyanotic congenital heart disease: does aspirin make a difference? Circulation 2007;116: 293–7. 5. Wessel DL, Berger F, Li JS, et al, CLARINET Investigators. Clopidogrel in infants with systemic-to-pulmonary-artery shunts. N Engl J Med 2013;368:2377–84. 6. Heidari-Bateni G, Norouzi S, Hall M, Brar A, Eghtesady P. Defining the best practice patterns for the neonatal systemicto-pulmonary artery shunt procedure. J Thorac Cardiovasc Surg 2014;147:869–73. 7. Gedicke M, Morgan G, Parry A, Martin R, Tulloh R. Risk factors for acute shunt blockage in children after modified Blalock-Taussig shunt operations. Heart Vessels 2010;25: 405–9. 8. Guzzetta NA, Foster GS, Mruthinti N, Kilgore PD, Miller BE, Kanter KR. In-hospital shunt occlusion in infants undergoing a modified Blalock-Taussig shunt. Ann Thorac Surg 2013;96: 176–82. 9. Jacobs JP, Mayer JE Jr, Mavroudis C, et al. The Society of Thoracic Surgeons Congenital Heart Surgery Database: 2016 update on outcomes and quality. Ann Thorac Surg 2016;101: 850–62. 10. The Society of Thoracic Surgeons. STS Congenital Heart Surgery Database Data Specifications. Version 3.2. Available at: http://www.sts.org/sites/default/files/documents/ CongenitalDataSpecsV3_22.pdf. Accessed January 9, 2017. 11. Alkhulaifi AM, Lacour-Gayet F, Serraf A, Belli E, Planche C. Systemic pulmonary shunts in neonates: early clinical outcome and choice of surgical approach. Ann Thorac Surg 2000;69:1499–504. 12. Myers JW, Ghanayem NS, Cao Y, et al. Outcomes of systemic to pulmonary artery shunts in patients weighing less than 3 kg: analysis of shunt type, size, and surgical approach. J Thorac Cardiovasc Surg 2014;147:672–7. 13. Williams JA, Bansal AK, Kim BJ, et al. Two thousand BlalockTaussig shunts: a six-decade experience. Ann Thorac Surg 2007;84:2070–5. 14. Fenton KN, Siewers RD, Rebovich B, Pigula FA. Interim mortality in infants with systemic-to-pulmonary artery shunts. Ann Thorac Surg 2003;76:152–6.
8
RICHARD E. CLARK AWARD EARLY SHUNT FAILURE
DO ET AL
15. Petrucci O, O’Brien SM, Jacobs ML, Jacobs JP, Manning PB, Eghtesady P. Risk factors for mortality and morbidity after the neonatal Blalock-Taussig shunt procedure. Ann Thorac Surg 2011;92:642–51. 16. Ohye RG, Schonbeck JV, Eghtesady P, et al. Cause, timing, and location of death in the Single Ventricle Reconstruction trial. J Thorac Cardiovasc Surg 2012;144:907–14. 17. O’Connor MJ, Ravishankar C, Ballweg JA, et al. Early systemic-to-pulmonary artery shunt intervention in
Ann Thorac Surg 2017;-:-–-
neonates with congenital heart disease. J Thorac Cardiovasc Surg 2011;142:106–12. 18. Emani S, Zurakowski D, Baird CW, et al. Hypercoagulability panel testing predicts thrombosis in neonates undergoing cardiac surgery. Am J Hematol 2014;89:151–5. 19. Emani S, Trainor B, Zurakowski D, et al. Aspirin unresponsiveness predicts thrombosis in high-risk pediatric patients after cardiac surgery. J Thorac Cardiovasc Surg 2014;148: 810–4; discussion 814–6.