Author + information
- Received June 1, 2005
- Revision received October 4, 2005
- Accepted October 10, 2005
- Published online April 4, 2006.
- Kerstin M. Amark, MD⁎,
- Tara Karamlou, MD†,
- Aoife O’Carroll, MS‡,
- Cathy MacDonald, MD‡,
- Robert M. Freedom, MD‡,
- Shi-Joon Yoo, MD§,
- William G. Williams, MD†,
- Glen S. Van Arsdell, MD†,
- Christopher A. Caldarone, MD† and
- Brian W. McCrindle, MD, MPH‡,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Brian W. McCrindle, The Hospital for Sick Children, 555 University Avenue, Toronto, Ontario, Canada M5G 1X8
Objectives We described morphologic characteristics, particularly pulmonary anatomy, and determined the prevalence of definitive end states and their determinants in children with pulmonary atresia associated with ventricular septal defect (PAVSD).
Background Pulmonary atresia associated with ventricular septal defect represents a broad morphologic spectrum that greatly influences management and outcomes.
Methods From 1975 to 2004, 220 children with PAVSD presented to our institution. Blinded angiographic review (n = 171) characterized bronchopulmonary segment arterial supply.
Results A total of 185 patients underwent surgery, and repair was definitive in 75%. Initial operations included systemic-pulmonary artery shunt in 57%, complete primary repair in 31%, or right ventricular outflow tract reconstruction in 12%. Based on angiographic review, 118 patients had simple PAVSD and 53 patients had PAVSD with major aortopulmonary collateral arteries (MAPCAs). Overall survival from initial operation was 71% at 10 years. Risk factors for death after initial operation included younger age at repair, earlier birth cohort, fewer bronchopulmonary segments supplied by native pulmonary arteries, and initial placement of a systemic-pulmonary artery shunt. Competing-risks analysis for initially palliated patients predicted that after 10 years, 68% achieved complete repair (with associated factors including later birth cohort and more bronchopulmonary segments supplied by native pulmonary arteries), 22% died without repair, and 10% remained alive without repair. Reoperations after complete repair occurred in 38 children (27%), with risk factors including older age at palliation, MAPCAs, and more segments supplied by collaterals.
Conclusions Outcomes in children with PAVSD have improved over time, and are better in completely repaired cases. Bronchopulmonary arterial supply is an important determinant of mortality, achievement of definitive repair, and post-repair reoperation.
The complexity of bronchopulmonary anatomy complicates management of patients with pulmonary atresia and ventricular septal defect (PAVSD) (1–5). Early primary repair has been applied in patients with “simple” PAVSD (well-developed central pulmonary arteries supplied by a ductus arteriosus), analogous to the treatment paradigm for tetralogy of Fallot (6). Several groups have similarly advocated single-stage complete unifocalization in patients with major aortopulmonary collateral arteries (MAPCAs) (7,8), but important interim attrition and morbidity with this approach mandates further evaluation. Management of patients with PAVSD and MAPCAs has evolved at the Hospital for Sick Children from a multi-staged approach to one emphasizing complete single-stage unifocalization in patients with adequate central pulmonary arterial size.
We sought to describe morphologic characteristics, with special attention to pulmonary anatomy, and to determine the prevalence of definitive end states and their determinants in children with PAVSD. We also sought to characterize outcomes over time and to establish whether trends could be correlated with the predominant management strategy operational in that era.
After approval by the Research Ethics Board at the Hospital for Sick Children, 220 children ≤18 years of age with unrepaired PAVSD referred from 1975 to 2004 were identified by the cardiology database and their medical records were reviewed (Table 1).Data were collected from admission, before any intervention, and at the last available follow-up. Patients or caregivers were contacted by telephone in cases in which current follow-up data were unavailable. Median follow-up time was 3.8 years (ranging to 26 years) from first admission, and is 98% complete for survivors.
Pulmonary arterial anatomy
Pulmonary arterial anatomy was documented by single-observer blinded review of the initial (n = 171) and follow-up (n = 98) angiograms (Table 1). Echocardiographic data were used for determination of intracardiac anatomy in patients without cardiac catheterization (n = 49). Central pulmonary arteries were located angiographically by relation to main bronchi, or by the seagull sign (9). Pulmonary arteries and the aorta were measured in the same angiogram or cardiac cycle phase. Non-dichotomous unpredictable peripheral branching was considered abnormal, as were dilated and tortuous pulmonary arteries. Stenoses were graded as present or absent. Direct collaterals were defined as native vessels arising from the aorta, and indirect collaterals were defined as originating from major aortic branches (10). Based on initial angiographic findings, the supply to the 18 bronchopulmonary segments was defined using the following definitions: 1) isolated segments had no connection to the central pulmonary arteries; 2) dual supplied segments were supplied by native pulmonary arteries via the ductus arteriosus and collaterals; 3) unperfused segments had no arterial supply; and 4) abnormal segments were defined qualitatively as those with an abnormal peripheral branching pattern. Patients with fully characterized pulmonary arterial anatomy (n = 171) were then classified into two groups based on whether the majority of the bronchopulmonary segments were supplied by collaterals, the MAPCA group (n = 53), or by native pulmonary arteries through the ductus arteriosus, the non-MAPCA group (n = 118). In follow-up angiograms, hemodynamic measurements were recorded, and right ventricular dilation, right and left ventricular function, and pulmonary regurgitation were qualitatively described (Table 2).
Data are presented as frequency, median with range, or mean ± SD as appropriate, with the number of missing values indicated. Percentages, hazard functions, and competing risk estimates are presented with confidence limits equivalent to 1 standard error (68%). All data analyses were performed using SAS statistical software (version 9, SAS Institute, Cary, North Carolina). Categorical and continuous non-timed event outcomes were evaluated with multivariable logistic and linear regression analyses. Multiphase parametric modeling of the hazard function (11) and competing risks methodology (12,13) were used to determine rates of transition to mutually exclusive time-related events. Multiphase parametric modeling (14) is a totally parametric (as opposed to the semi-parametric method established by Cox ) method of analysis that accommodates both the time-related freedom from an event and also the time-varying nature of risk (phases of risk) for that event. Hazard function analysis that incorporates these phases is ideal for understanding many postoperative events, including those in the present study, in which the early risk of an event such as death (i.e., hazard) after surgery is initially high, rapidly declines to a more constant non-zero hazard, followed by a later increase in risk thereafter.
The graphic depiction of overall survival generated from modeling the hazard function (which in this case has two phases) contains solid lines, which are continuous point estimates enclosed by 70% confidence limits. Model building begins by nonparametric (Kaplan-Meier) estimation, which reveals the shape of risk over the period of follow-up (i.e., is there early attrition followed by a gradual increase, or is there only ongoing constant risk?) Hazard models are then constructed using the log-likelihood method, and model fit is also examined graphically by comparing the parametric estimates to the nonparametric estimates (shown in the graph as superimposed circles with error bars).
Competing risks analysis is a method of time-related data analysis in which multiple, mutually exclusive events are considered simultaneously. Competing risks analysis is integrative in that it considers multiple outcomes in the context of one another, and can therefore address the question of how often an event may occur in the presence of other events for which a patient is at simultaneous risk. A familiar example of this phenomenon is the estimated prevalence of prosthetic valve replacement adjusted for the estimated prevalence of death occurring before replacement. The present study considers that patients make a transition from an initial state (called event-free survival after palliation) to two other states (complete repair and death without complete repair) that are considered to be terminating. All palliated patients in this example begin alive at time zero, and thereafter migrate (or transition) to the two specified end states at a rate determined by the underlying hazard function. Rates of transition, or rates of migration, from the initial state to one of the events (called an end state) are individual, independent hazard functions. Incremental risk factors associated with each state were identified by multivariable regression analysis as previously described (16). Mathematical transformations of continuous variables, such as logarithms, polynomials, and square or inverse functions, were used to optimize calibration of the variable to the risk of outcome events, and interactions among retained variables in the model were considered in all multivariable analyses. Variable selection was guided by bootstrap validation (16).
Initial morphologic characteristics
Bronchopulmonary anatomy is characterized by a continuous morphologic spectrum, but as confirmation of our definition, MAPCA patients are polarized at either end of this spectrum (Fig. 1).Compared with those in the non-MAPCA group (n = 118), the 53 MAPCA patients had a smaller McGoon ratio (1.0 ± 0.4 vs. 1.4 ± 0.5), fewer bronchopulmonary segments supplied by the true pulmonary arteries (5.0 ± 0.6 vs. 16.0 ± 0.3), and more segments with a dual supply (9.0 ± 0.9 vs. 1.0 ± 0.3) or that were unperfused (2.0 ± 0.3 vs. 0.0 ± 0.1) (p < 0.001 for all).
Initial surgical procedures
Of the initial 220 patients, operative intervention was undertaken at first admission in 187 at a median age of 18 days (range, 0 to 10.6 years), 2 of whom had exploration only without any reparative procedure (Fig. 2).Systemic-pulmonary artery shunts included modified Blalock-Taussig shunts and end-to-side (“Mee”) shunts. Right ventricular outflow tract (RVOT) reconstruction included the following procedures: surgical valvotomy with or without concomitant transannular patching, transannular patch placement, or interposition of a conduit between the right ventricle and the pulmonary artery. Two patients late in the series had radiofrequency-assisted valvotomy and balloon dilation. Initial procedures were performed in 36 patients whose angiograms were unavailable for blinded angiographic review, including systemic-pulmonary artery shunt in 20, primary complete repair in 14, and RVOT reconstruction in 2 patients.
Non-MAPCA patients (n = 118)
A flowchart of events in patients according to diagnostic group is shown in (Fig. 3).Four patients without MAPCAs, but with other severe non-cardiac anomalies, had no operative intervention. Palliation was performed in 83 and primary complete repair, defined as VSD closure and RVOT reconstruction, occurred in 31 patients.
MAPCA patients (n = 53)
Eighteen patients with MAPCAs had no operative intervention. The MAPCA patients were more likely to have non-operative management (odds ratio [OR], 14.7; 95% confidence interval [CI], 4.6 to 46.2; p < 0.001) Palliation was performed in 23 patients, including RVOT reconstruction in 13 and systemic-pulmonary artery shunt in 10 patients. Procedures directed at MAPCAs concomitant with RVOT reconstruction included complete unifocalization in 5 and partial unifocalization and ligation in 1 patient each. Systemic-pulmonary artery shunts occurred simultaneously with partial unifocalization in 7 patients. Single-stage complete repair, defined as VSD closure, RVOT reconstruction, and single-stage complete unifocalization, occurred in 12 patients.
Evolution of initial surgical management over time
Primary complete repair (n = 57) was strongly associated with more recent era of operation for all patients regardless of pulmonary arterial anatomy, with 67% (n = 38) of initial repairs occurring in era 3 (1995 to 2004), compared with only 32% (n = 18) in era 2 (1985 to 1994), and only 2% (n = 1) in era 1 (1975 to 1984) (p < 0.0001) (Fig. 4).In the MAPCA group, 83% (n = 10) of complete primary repairs occurred era 3, compared with only 17% (n = 2) in era 2, and none in era 1 (p < 0.0001). Similarly, in the non-MAPCA group, 97% (n = 30) of initial repairs occurred in era 2 and 3, compared with 3% (n = 1) in era 1 (p < 0.0001).
Initial catheter interventions
Initial procedures occurred in 23 patients before repair, including balloon dilatation of branch pulmonary arteries ± stent placement in 12, collateral coiling in 7, and pulmonary valve perforation in 4 patients.
Overall time-related survival from initial operation was 71% at 10 years (Fig. 5).Mortality rate after initial operation was highest during the first year after operation and tapered thereafter. Incremental risk factors for time-related death after operation were sought and are listed in Table 3.Thirty deaths (25%) occurred in patients without MAPCAs, and 21 occurred (40%) in the MAPCA group (p = 0.06).
Achievement of definitive repair
Complete repair occurred in 139 patients at a median age of 2 years (range, 0 to 15 years). Primary repair without prior palliation occurred in 57 patients. Staged repair was accomplished in 82 patients, 63 in the non-MAPCA group, 9 in the MAPCA group, and 10 in patients without angiographic review. Competing risks analysis for the 128 patients who underwent initial palliation predicted that at 10 years after palliation, 68% had achieved complete repair, 22% had died without complete repair, and 10% remained alive without repair (Fig. 6).
Factors increasing transition rates to complete repair in the 185 who underwent initial reparative operation included later birth cohort and more bronchopulmonary segments supplied by the true pulmonary arteries (Table 3). Decreased transition rates to complete repair and increased pre-repair attrition are evident in patients with MAPCAs, as shown in Figure 7.
Outcomes after complete repair
A total of 38 children underwent 47 reoperations after complete intracardiac repair (Table 4).For the 139 patients who underwent complete repair, competing risks analysis predicted that at 5 years after complete repair, 60% remained alive without subsequent operation, 28% had a second operation, and 12% had died without reoperation. Incremental risk factors for reoperation after complete intracardiac repair included being in the MAPCA group, older age at initial palliation, and a greater number of collateral segments (Table 3).
After complete intracardiac repair, 104 percutaneous interventions occurred in 56 patients at a median age of 1.1 years (range, 0 to 13 years) from repair, including balloon dilatation of branch pulmonary arteries ± stent placement in 75, balloon dilation ± stent placement of existing RVOT conduit in 22, and collateral coiling in 7 patients. Percutaneous interventions were more prevalent in the MAPCA group (25%) than in the non-MAPCA group (17%), and at a decreased interval from complete repair (1 year vs. 4 years, p = 0.003). Competing risks analysis predicted that at 3 years after complete repair, 35% underwent catheter-based reintervention, 6% had died without reintervention, and 59% remained alive without reintervention. Risk factors for reintervention included use of a homograft for RVOT reconstruction, older age at complete repair, and single-stage complete repair in MAPCA patients (Table 3). A significant interaction existed, however, between complete repair in the MAPCA group and age at repair, showing that earlier single-stage repair increased the risk of subsequent interventions (Fig. 8).
Clinical follow-up from first admission was obtained in 216 patients at a median interval of 3.8 years. Mean oxygen saturation (%) at follow-up was higher in patients who underwent primary complete repair (96 ± 3) versus those who had staged palliation (90 ± 12, p < 0.001). Follow-up angiograms were obtained in 98 survivors (62%) at a median interval of 6 years from initial admission (Table 2). Mean right ventricular/systemic pressure (%) was 55 ± 20. Asymptomatic clinical status was associated with primary complete repair (OR, 3.6; 95% CI, 1.2 to 10.9; p = 0.02), achievement of complete repair at any time (OR, 13.0; 95% CI, 3.7 to 47; p < 0.001), and the absence of MAPCAs (OR, 0.3; 95% CI, 0.1 to 0.8; p = 0.01) on multivariable analysis after adjustment for follow-up duration and operation era. At 6 years after initial operation, patients who underwent systemic-pulmonary artery shunts had less pulmonary arterial growth (median Δ Nakata = 74; range, 133 to 430) than those undergoing other initial operations after correction for initial anatomy (median Δ Nakata = 102; range, 205 to 520), p = 0.08.
The 10-year survival in our series of 71% is higher than in other historical series (3,17,18). Incremental risk factors for death after initial operation in this study included earlier birth cohort, systemic-pulmonary artery shunt placement, increasing number of lung segments supplied by MAPCAs, and younger age at complete repair. Improved survival for those undergoing operation in a later era is multifactorial.
The prevalence of primary complete repair increased dramatically in the later eras, as did the proportion of patients reaching definitive repair. Similarly, the use of systemic-pulmonary artery shunts as the first stage decreased with later birth cohort in favor of RVOT reconstruction. Patients undergoing systemic-pulmonary artery shunts had increased mortality independent of other factors, potentially related to decreased pulmonary parenchymal recruitment with this technique (5,19,20). Decreased pulmonary arterial growth rate, measured by Δ Nakata index, in patients with shunts supports this contention. An earlier report by Freedom et al. (10) from our institution describing our initial experience with RVOT reconstruction showed no conclusive benefit with respect to symmetric pulmonary arterial growth, but included only 15 patients, only 5 of whom had angiographic follow-up. The potential disadvantage of systemic-pulmonary artery shunts has been suggested by the results from other series (5,19,20). Other factors not evaluated in this study likely also contributed to improved results over time.
Variables relating to the morphology of the pulmonary arterial circulation are well-known risk factors for death after repair, but their influence on other important outcomes, including the achievement of complete repair, reoperations, percutaneous reintervention, and long-term clinical status, has not been previously described (7,21–24). We have fully characterized the bronchopulmonary anatomy in the largest series of patients with PAVSD, and have shown that increased collateral supply adversely impacts end-state achievement independent of choice of initial management. This risk factor is likely a surrogate for elevated pulmonary arterial or vascular resistance, which is inversely related to the number of pulmonary arterial segments connected to an ipsilateral central pulmonary artery (25). The importance of evaluating pulmonary arterial supply not only before initial treatment but also throughout the patient’s clinical course is highlighted by the profound influence of bronchopulmonary arterial anatomy on outcomes (22).
We were surprised that given the known advantages of early surgical intervention and the improved long-term clinical status in repaired cases, earlier age at repair was a risk factor for death after operation. One potential explanation is that median age at repair in our series is lower than in many other reports despite the large preponderance of staged patients (3,5,8,17,18,22,23,26). Thus, the finding of early repair as a risk factor may reflect the fact that the designation “early” is potentially arbitrary and depends on one’s perspective. More likely, however, is that commitment to early definitive repair is a long-term investment requiring some initial cost. This is analogous to the evolution in management favoring arterial switch operation over atrial switch operation, in which the historically slightly higher mortality of a new operation was mitigated by the potential long-term functional benefit (27). Patients who underwent repair and survived had greater improvement in long-term clinical status, were less cyanosed, and had more favorable pulmonary arterial characteristics at follow-up compared with those who remained palliated or did not undergo surgery.
Achievement of complete repair
Achievement of complete repair occurred in 70% of patients in our study, which was similar to the rates achieved in other reports (3,17,18,22). Later year of operation was strongly associated with definitive repair because of the increased prevalence of primary repair and because enough time elapsed for previously staged patients to reach this state. Patients who had undergone repair had acceptable mean right ventricular/left ventricular pressures at late follow-up and significant growth of the pulmonary arterial vascular bed.
Reoperation after definitive repair occurred frequently and was associated with increased number of MAPCAs and later age at initial palliation. Delayed palliation increases the duration for which the collateral vessels are exposed to systemic arterial pressure and may predispose to the development of pulmonary vascular obstructive disease in supplied lung segments, especially in the absence of important MAPCA stenoses (8). Encouraging, however, is that despite the high incidence of post-repair reoperation, the majority of non-MAPCA survivors were asymptomatic and clinically well.
The MAPCA patients undergoing single-stage complete repair had an increased risk of reintervention, and this effect was more pronounced when repair was performed at an earlier age. High reintervention risk in this group may contribute to our finding that completely repaired MAPCA patients had a worse clinical status than completely repaired non-MAPCA patients. Additionally, catheter-based intervention accounted for six deaths in our series. Risk rates of reintervention in the MAPCA group approximated those in the non-MAPCA group after 8 months of age, suggesting that delaying single-stage repair outside of the neonatal period may afford some benefit. Although it can be argued that sequential reinterventions in patients with MAPCAs are mandated as part of the treatment algorithm, and therefore may not represent adverse events per se, clear evidence supporting this notion is currently lacking.
We have shown important improvement in outcomes over time that are associated with a shift in treatment paradigm emphasizing primary repair for those with simple PAVSD and an individualized approach for those with MAPCAs tailored to the adequacy of the pulmonary arterial bed. Our results confirm that RVOT reconstruction as the initial procedure is preferable to systemic-pulmonary artery shunt placement for patients in whom primary repair is not feasible, and that surgical decisions should be tailored to well-defined bronchopulmonary anatomy.
- Abbreviations and Acronyms
- major aortopulmonary collateral arteries
- pulmonary atresia associated with ventricular septal defect
- right ventricular outflow tract
- Received June 1, 2005.
- Revision received October 4, 2005.
- Accepted October 10, 2005.
- American College of Cardiology Foundation
- Metras D.,
- Chetaille P.,
- Kreitman B.,
- Fraisse A.,
- Ghez O.,
- Riberi A.
- Bull K.,
- Somerville J.,
- Ty E.,
- Spiegelhalter D.
- Carotti A.,
- Albanese S.B.,
- Minniti G.,
- Guccione P.,
- DiDonato R.M.
- Reddy V.M.,
- McElhinney D.B.,
- Amin Z.,
- et al.
- Tchervenkov C.I.,
- Salasidis G.,
- Cecere R.,
- et al.
- Jefferson K.,
- Rees S.,
- Somerville J.
- McGriffin D.C.,
- Naftel D.C.,
- Kirklin J.K.,
- et al.,
- Pediatric Heart Transplant Study Group,
- Pediatric Heart Transplant Study Group
- ↵The Cleveland Clinic Heart and Vascular Institute. Hazard Function Technology. Available at: http://www.clevelandclinic.org/heartcenter/hazard. Accessed October 1, 2004.
- Cox D.R.
- Yagihara T.,
- Yamamoto F.,
- Nishigaki K.,
- et al.
- Culbert E.L.,
- Ashburn D.A.,
- Cullen-Dean G.,
- et al.