clinical outcomes after the arterial switch operation for

J W Kirklin, E H Blackstone, C I Tchervenkov and A R Castanedaprocedural, and institutional risk factors. Congenital Heart Surgeons Society.

Clinical outcomes after the arterial switch operation for transposition. Patient, support,

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright © 1992 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation doi: 10.1161/01.CIR.86.5.1501

1992;86:1501-1515Circulation.

http://circ.ahajournals.org/content/86/5/1501the World Wide Web at:

The online version of this article, along with updated information and services, is located on

http://circ.ahajournals.org//subscriptions/

is online at: Circulation Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. Permissions and Rights Question and Answer information about this process is available in the

located, click Request Permissions in the middle column of the Web page under Services. FurtherEditorial Office. Once the online version of the published article for which permission is being requested is

can be obtained via RightsLink, a service of the Copyright Clearance Center, not theCirculationpublished in Requests for permissions to reproduce figures, tables, or portions of articles originallyPermissions:

at HOSPITAL FOR SICK CHILDREN on September 4, 2013http://circ.ahajournals.org/Downloaded from

http://circ.ahajournals.org/content/86/5/1501

http://www.ahajournals.org/site/rights/

http://www.lww.com/reprints

http://circ.ahajournals.org//subscriptions/

http://circ.ahajournals.org/

1501

Clinical Outcomes After the Arterial SwitchOperation for Transposition

Patient, Support, Procedural, and Institutional Risk Factors

John W. Kirklin, MD; Eugene H. Blackstone, MD; Christo I. Tchervenkov, MD;Aldo R. Castaneda, MD; and the Congenital Heart Surgeons Society

Background. As the probability increases that the arterial switch operation is optimal treatment fortransposition, detailed information about outcomes and the circumstances in which they are suboptimalbecomes important.Methods and Results. A multi-institutional prospective study with annual detailed follow-up included 513

neonates with simple transposition or transposition and ventricular septal defect entering for diagnosisand treatment at <15 days of age and undergoing an arterial switch repair. The 1-month and 1- and5-year survivals were 84%, 82%, and 82%, respectively. The hazard function for death had a rapidlydeclining single phase that approached zero by 12 months after surgery. Among the eight patients whodied >3 months after the operation, four had severe ventricular dysfunction, probably related to imperfectcoronary arterial transfer. Coexisting single ventricular septal defect was not a risk factor for death.Origin of the left main coronary artery or only the left anterior descending or the circumflex artery fromthe right posterior sinus (sinus 2) was a risk factor that was even stronger when an intramural course was

present; multiplicity of ventricular septal defects was a risk factor. Longer global myocardial ischemictime and total circulatory arrest time were risk factors. Certain institutions were shown to be risk factorsfor death; the results in some improved with increasing experience, in some they did not, and in some theyworsened.

Conclusions. Good early and intermediate-term clinical outcomes can be obtained in neonates withsimple transposition and transposition and ventricular septal defect by use of the arterial switchoperation. Certain coronary artery patterns and certain institutions lessen the goodness of outcome.(Circulation 1992;86:1501-1515)KEY WoRDs * congenital heart disease * mortality

Asthe probability increases that the arterial switchoperation is optimal treatment for patients withtransposition, detailed information about out-

comes and the circumstances (risk factors) in which theyare suboptimal becomes more important. A multi-institutional study with annual detailed follow-up pro-vides a basis for analyses directed toward obtaining thisinformation in a manner that will be useful in widelydiffering circumstances.

MethodsPatientsA total of 895 neonates 1) with complete transposi-

tion of the great arteries (TGA) and 2) admitted to aninstitution before 15 days of age were entered prospec-tively between January 1, 1985, and March 1, 1989, intoa multi-institutional study conducted by the Congenital

From the Division of Cardiothoracic Surgery (J.W.K., E.H.B.),The University of Alabama at Birmingham Medical Center; theDivision of Cardiovascular and Thoracic Surgery (C.I.T.), McGillUniversity, Montreal; and the Department of Cardiac Surgery(A.R.C.), The Children's Hospital, Boston.Address for reprints: John W. Kirklin, MD, University of

Alabama at Birmingham, University Station, Birmingham, AL35294.

Received November 27, 1991; revision accepted August 5, 1992.

Heart Surgeons Society (CHSS). Treatment was notrandomly assigned or directed by the study but ratherwas selected by the physicians caring for the patients onthe basis of their knowledge and experience. Entry into

See p 1654the study is now closed, but follow-up continues. Cer-tain aspects of the experience have been presentedpreviously, as have criteria for the diagnosis of thevarious subsets.1-3An arterial switch repair was chosen as therapy by the

patient's physician and was done in 513 of the 846patients with simple transposition or TGA and ventric-ular septal defects (VSD). An atrial switch repair waschosen in 277 patients. Nine other patients underwent adifferent repair. Twelve additional patients were alivebut as yet unrepaired at last follow-up. Thirty-fivepatients had died without repair.Among the 384 patients with simple TGA undergoing

an arterial switch repair, more than one third (37%)entered into the study within 24 hours of birth, andamong the 129 with TGA and VSD undergoing thisrepair, 32% had entered by that time (see "AppendixA," Figure A-1). The median birth weights in each ofthese groups were 3.4 and 3.5 kg, respectively, butvaried widely in both groups (Figure A-2). The coexist-



1502 Circulation Vol 86, No 5 November 1992

TABLE 1. Coronary Arterial Pattern in Patients Undergoing the Arterial Switch Operation for Simple TGA andTGA With VSD

Origin of coronary arteries Percent of nTotal deaths

Sinus 1 Sinus 2 n (513) (%) No. %LADCx R 367* 72 60 16LAD R 1 0.2 0 0LADCxR ... 5 1 1 20... LADCxR 36 7 10 28

Single ostium, or twoclosely placed ostiat 26 5 7 27Two ostia, oneeccentric to the leftand near commissuret 10 2 3 30

LADR Cx 13 3 3 23LADR LADCx 2 0.4 1 50R LADCx 8 2 3 38LAD CxR 75 15 14 19RvLADCx CxR 1 0.2 0 0Malaligned aortic commissures§ 2 0.4 0 0Unknown 3 1 0 0

Total 513 92 18TGA, transposition of great arteries; VSD, ventricular septal defect; LAD, left anterior descending coronary artery;

Cx, circumflex coronary artery; R, right coronary artery; Rv, conal (right ventricular) branch of right coronary artery.*Three of the 367 (no deaths) had malaligned pulmonary commissures.tLAD or left coronary artery passed between aorta and pulmonary artery (intramural) in two (one death) of the 26.*All passed between aorta and pulmonary artery (intramural).§In one, all coronaries arose from a posterior facing sinus; in one, LADCx arose from a posterior facing sinus, R from

a right-sided sinus.

ing cardiac and noncardiac malformations in these 513patients undergoing the arterial switch operation aretabulated in Tables A-1 and A-2. The age at operationwas <7 days in 63% of the patients with simple TGAand in 36% of those with TGA and VSD (Figure A-3).

Coronary Arterial AnatomyOf the patients undergoing the arterial switch repair,

72% had the left anterior descending and circumflexcoronary arteries arising from a single orifice in sinus 1(left posterior facing sinus) and the right coronaryartery arising from sinus 2 (right posterior facing si-nus).4 Most of the other variations described as occur-ring in patients with TGA were encountered in thisexperience (Table 1), including an intramural pathwayof the left main or left anterior descending artery in 12patients.5

Relations of the Great ArteriesMost commonly, the aorta was directly anterior, less

commonly it lay anteriorly but somewhat to the right,and uncommonly it lay anteriorly but somewhat to theleft (Table A-3). The very uncommon variant in whichthe aorta lies posterior to the pulmonary trunk was notrepresented in this group of patients.6

OperationIn addition to the 513 patients who actually received

an arterial switch repair, 22 patients went to surgerywith the intent that an arterial switch operation wouldbe performed, but the surgeon decided not to perform itafter viewing the morphology. In 14 (four of whom

died), a Senning atrial switch repair was done instead;in two (who survived), an atrial septectomy was donefollowed later by a Senning repair; in one (who sur-vived), a previously placed systemic-pulmonary arteryshunt was narrowed and a Senning repair performedlater; in one other (who died), a pulmonary trunk bandwas placed and an atrial septal defect was enlarged; andin three patients (none of whom died), no procedurewas performed at that time, and later, a Senning repairwas done. One other patient had placement of a pul-monary trunk band and later became one of the 513patients who underwent the arterial switch operation(he died after the arterial switch operation). In 280patients, the arterial switch repair was done during along period (.30 minutes) of profoundly hypothermictotal circulatory arrest, and in 233, it was done undercardiopulmonary bypass with or without shorter periodsof total circulatory arrest. Information as to myocardialmanagement was available in 475 of the 513 patients;cold cardioplegia was used in 474 of those 475 patients.Information about the cardioplegic agent and vehiclewas not available for most of the patients. The details ofthe procedure itself varied considerably betweeninstitutions.

Follow-upAll patients (n=895) entered into the study and not

known to be dead (n=720) were traced between May 1and August 1, 1990, either by the Data and AnalysisCenter (n=574) or by the parent institution (n=146).An inquiry letter with a form was sent initially to theparents (or guardians) of each patient; when no answer



Kirklin et al Risk Factors in Arterial Switch Operation for Transposition 1503

was received within 3 weeks, strenuous attempts weremade to locate and interrogate, by telephone, the familyor other responsible persons. After these efforts, infor-mation was still not obtained during this follow-up in 26patients not known to be dead. Among these, the lastsuccessful follow-up was in 1987 in two, in 1988 in eight,and in 1989 in 16. The analyses used the generallyaccepted technique of censoring incompletely tracedpatients after the time of last follow-up.Among the 895 patients, the median interval between

entry and follow-up was 36.5 months (1.3-65.6 months).The mean follow-up interval was 37.5+14.16 months.The intervals in the 513 patients undergoing the arterialswitch operation were entirely similar.

Data Collection and AnalysisFor each patient, copies of the pertinent parts of the

hospital chart for each admission, including reports ofdiagnostic and imaging studies, of the operations, and ofany autopsy as well as summaries of the preoperativeand postoperative course, were forwarded to the Dataand Analysis Center at the University of Alabama inBirmingham. Here they were assembled and retained;they were also entered and stored in a confidential filein an IBM 9370 minimainframe computer. Confidenti-ality of all records, data, and analyses was rigorouslymaintained.The usual explorations of the data and relations were

made by cumulative distribution plots, contingency ta-bles, and life table methods.7 Time-related freedomsfrom death (death after entry, death after repair) werealso studied by hazard function regression analysis, andin this domain multivariable analyses, includingsearches for interaction terms, were performed andinternal validations obtained.8 Multiple correlations be-tween variables were made and studied.

Multivariable AnalysesTwo series of multivariable analyses were made using

a probability value ofp<0.1 as the criterion for includ-ing the variable in the final equation. This value ofp waschosen to avoid overlooking a possibly important riskfactor in a study with admittedly suboptimal statisticalpower. The series is by far the largest reported series,however, and is the only prospective study of thiscondition.

In one series of analyses, four parsimoniously derivedequations were obtained after entering 1) only patient-specific variables, 2) patient-specific variables and avariable representing arbitrarily assembled high-riskand low-risk groups of institutions along with markersfor experience, 3) patient-specific variables, the twogroups of institutions along with markers for experi-ence, and each institution individually, and 4) all of theprevious groups of variables plus procedural variables.Two markers for institutional experience were used: thenumber of arterial switch operations previously per-formed in the CHSS study by the institution and theinterval (in years) since the first arterial switch opera-tion for a CHSS patient in the institution. Variablesentered into all analyses are depicted in "Appendix B."

High-risk (n= 11) and low-risk institutions were de-fined and identified in an earlier study.3 At that time,two institutions had the lowest time-related percent of

TABLE 2. Total Deaths (as Percentage of the Number Undergoingthe Arterial Switch Operation) According to the Institution inWhich the Operation Was Performed*

Total deaths

InstitutiontABtCDEFtGHI

KLMN0p

QRSTUV

Confidence interval(%)

6779121213232933365080676767677000

100100

3-125-101-221-286-219-174-2917-3117-4518-5227-4718-8247-9738-8838-8824-9624-9649-860-610-8515-10015-100

Total§ 17.9 16.2-19.9

*Low-risk institutions (whose number of patients varied be-tween 11 and 146) are institutions A-G. High-risk institutions(whose number of patients varied between three and 52) areinstitutions H-R. Institutions S-V entered only one or two pa-tients into the study, and they are not considered individually inany subsequent analyses.tThe confidence intervals of institutions A-G are low and in all

instances overlap with each other. This is not true of institutionsH-V. Institutions H-L have confidence intervals that are alloverlapping one with another.

*These are the two largest experiences, which together total 235patients with 21 deaths (8.9%; confidence interval, 7.0-11.3%).§The total number of patients was 513, and the total number of

deaths was 92.

deaths, and the mortality (in the hospital and later) in acombined group consisting of the patients of these twoinstitutions was 9% (70% confidence limits, 6-13%).These two institutions, plus the five others whose 70%confidence limits around their percent mortality over-lapped confidence limits of 6-13%, were considered tobe low-risk institutions; all others were consideredhigh-risk. These same seven institutions had the lowestpercentages of total deaths in the present study, con-ducted several years later, and the confidence intervalsfor percent death of each of the seven institutions wereoverlapping with those of each of the other institutionsin the low-risk group (see Table 2). The validity of thisgrouping was further confirmed by the multivariableanalysis (see "Appendix D," Table D-2). These groupvariables (high-risk institutions and low-risk institu-tions) were entered into some analyses to have a

1 1%




TABLE 3. Incremental Risk Factors for Death After the ArterialSwitch Operation for Simple TGA and TGA with VSD*

Single hazardIncremental risk factors for death phase p

PatientLCA, LAD, or Cx arising from sinus 2

Low-risk institutions 0.08High-risk institutions <0.0001

Intramural course of LCA or LADt <0.0001Coexisting cardiac or noncardiac anomalies 0.002PA banding > 1 month previously* 0.0002

InstitutionsNumber of cases in low-risk institutions§ <0.0001

Institution Fil 0.05Number of cases in high-risk institutions¶ <0.0001

Institution K# 0.02Institution N** 0.006Institution Qtt 0.05Institution R* * 0.002

TGA, transposition of the great arteries; VSD, ventricularseptal defect; LCA, left coronary artery; LAD, left anteriordescending coronary artery; Cx, circumflex coronary artery; PA,pulmonary artery.

*Patient and institutional variables (high-risk vs. low-risk andexperience in terms of number of cases) and each individualinstitution were entered into this parsimoniously derived equation(n=513).

tThis variable was active only in high-risk institutions.*This variable was active only in low-risk institutions.§The following institution is an interaction term with "number

of cases." "Number of cases" has the same meaning as in Table 6.liThe coefficient is an increment (+ in this case, meaning that

the prevalence of death was greater along the slope and that thesurvivorship was less) to the slope for "all other low-risk"institutions.¶The following institutions are interaction terms with "number

of cases."#The coefficient is an increment (+ in this case) to the intercept

of "all other high-risk" institutions.**The coefficient is an increment (+ in this case) to the slope of

"all other high-risk" institutions.ttThe coefficient is an increment (- in this case) to the slope of

"all other" institutions.

sufficiently large number of events associated with thevariable to allow exploration of interaction terms (see,for example, Table 3).

In the second series of analyses, nonparsimoniouslyderived equations were obtained sequentially, begin-ning with one derived after only patient-specific vari-ables were entered; in these, variables retained in theequations earlier in the sequence were retained latereven when the probability value had become larger. Thevariables added for the subsequent sequential analyseswere 1) support techniques, 2) procedural techniques,3) individual institutions (by location), and 4) institu-tional experience (by both markers just described).Further sequential analyses were made after the groupsof variables entered were combined in various ways (seeTables D-1-D-5).

Shaping parameters, coefficients, and the variance-covariance matrices of the multivariable equations arenot reproduced in this publication but are availableupon request.

10(

gc

BC

D 42>4) (415..

60 82%

Interval (Months) ater Repair

0.10

0.09B

0.08 [0.07 1

:c

a.cU0

0.08

0.05

0.04

0.03

0.02

0.01

interval Hazard(months) (x 1000)

1/30 880

114 751 1212 0.5224 0.2236 0.1360 0.068

0.00 L .............,0 6 12 18 24 30 36 42 48 54 60 68

Interval (Months) after Repair

FIGURE 1. Graphs showing freedom from death after thearterial switch repair for simple transposition of the greatarteries (TGA) and TGA with ventricular septal defect. PanelA: Time-related percent survival. Each circle, positioned bythe life table method, represents a death, and the vertical barsrepresent the 70% confidence intervals of the actuarial esti-mate. The numbers in parentheses indicate the numbers ofpatients still at risk at the time of each estimate; this number(which is the number alive and traced) at 36 months, 48months, and 60 months after repair was 175, 71, and 12,respectively. The solid line represents the continuous pointestimate of the survivorship estimated by hazard functionregression analysis (as does the table), and the dashed linesenclose the 70% confidence intervals of the parametricestimate. Panel B: Hazard function for death.

ResultsSurvivalAmong the heterogeneous group of 513 patients with

simple transposition or TGA with VSD undergoing thearterial switch operation, the 1-month and 1- and5-year survivals were 84%, 82%, and 82%, respec-tively (Figure 1A). The hazard function for death hada rapidly declining single phase, which had becomenearly flat by 6 months after surgery and approachedzero by 12 months after surgery (Figure 1B). (After anew overall formal follow-up in July 1991, the percentsurvival and hazard functions were unchanged, butone patient was found to have died suddenly 40months after surgery.)Among the eight patients who died >3 months after

the arterial switch repair, four died 8.2-21 months after



Kirklin et al Risk Factors in Arterial Switch Operation for Transposition

TABLE 4. Incremental Risk Factors for Death After the ArterialSwitch Operation for Simple TGA and TGA With VSD*

Single hazardIncremental risk factors for death phase pLCA, LAD, or Cx arising from sinus 2Without an intramural course of LCA or LADt 0.05With an intramural course of LCA or LADt 0.02

Multiple VSDs 0.02Coexisting noncardiac anomalies 0.03PA banding > 1 month previously 0.09Older age at repairt 0.07

Simple TGA§ 0.08

TGA, transposition of great arteries; VSD, ventricular septaldefect; LCA, left coronary artery; LAD, left anterior descendingcoronary artery; Cx, circumflex coronary artery; PA, pulmonaryartery.

*Only patient variables were entered into the parsimoniousanalysis (n=513).tThese are mutually exclusive variables, each with its ownp and

coefficient, and are subsets of "LCA, LAD, or Cx arising fromsinus 2."

*This variable was active only in patients without PA banding>1 month previously.§Simple TGA, but not TGA with VSD, was a positive interac-

tion term with age at repair.

repair with severe ventricular dysfunction in all proba-bility related to imperfect coronary arterial transfer.One patient with simple TGA died 7.9 months afterrepair with severe pulmonary vascular disease.

Patient Incremental Risk Factors for DeathThe patient-specific risk factors for death after the

arterial switch repair are given in Table 4. The absenceof TGA with a single VSD as a risk factor is noteworthyand is consistent with the finding on life-table analysisthat the somewhat lesser time-related survival of thegroup of patients with TGA and VSD could have beencaused by chance alone (Figure 2).The only risk factor specifically related to the malfor-

mation of TGA was the coronary artery pattern. Theorigin of the left main coronary artery or only the leftanterior descending or circumflex coronary artery fromsinus 2 was a risk factor for death. An intramural courseof the left main or left anterior descending coronaryartery added further risk (Figure 3). Arteries with anintramural course nearly always arose from an ostiumnear the commissure between sinus 2 and sinus 1 (seeTable 1).

Multiplicity of VSDs (Figure 4) and important coex-isting noncardiac anomalies were risk factors for deathafter the arterial switch repair. Patients with a previ-ously placed band on the pulmonary trunk were atincreased risk of death after the arterial switch repairbut only if the band had been in place for more than 1month (see Table 4). Older age at surgery was a riskfactor for death after the arterial switch repair, partic-ularly when the age was >7 days (Figure 5).

Support Risk Factors for DeathUnivariably, the duration of profoundly total circula-

tory arrest was a risk factor for death (p=0.01). In mostof the multivariable analyses, the probability value ofpfor "myocardial ischemic time" was smaller than that

100

90

80

_ 70

30

260

1d50c

A. 40

30

20

I0

(3271(323) (318)....5.) ..................... ......

-- - - - - - - - - - -

j I jf~~Simple TGA withTGA VSDInterval jo n-384) (oi, nw129)

(monts) Life Para- Life Para-Table metric Table metric

1122460

87% 85% 81% 8U%84% 84% 79% 78%83% 83% (79%) 78%(83%) 83% (79%) 77%

P(Gehan-Wilcoxon) 0.3P(parametric) a 0.4

0 6 12 18 24 30 36 42 48 54Interval (months) after Arterial Switch Repair

60 66

FIGURE 2. Graph showing survival after the arterial switchrepair according to whether the patient had simple transposi-tion of the great arteries (TGA) or TGA with ventricularseptal defect (VSD). As in Figure 1, the circles and squaresand the vertical bars represent actuarial estimates based onactual experience. In contrast to Figure 1, the solid anddashed lines represent the time-related average of the risk-adjusted percent survival (and confidence intervals) of eachpatient with simple TGA or TGA with VSD obtained using thepatient-specific multivariable riskfactor equation (Table C-i).The method was used by Ferrazzi et al.9 In the tabulardepictions, "Life Table" refers to the actual Kaplan-Meierdepiction; "Parametric" refers to the average risk-adjustedpercent survival obtained as described. The close correspon-dence of the two constitutes an internal validation of theparametric method of risk adjustment, especially becauseneither simple TGA nor TGA with VSD was a risk factor inthe multivariable equation. The p values refer to the differencein actuarial percent survival between simple TGA and TGAwith VSD.

for "duration of profoundly hypothermic total circula-tory arrest"; when "duration of myocardial ischemictime" was applied only to patients in whom there was no

100

90

80o

- 70

50

43020

10

00

Co ar\,ay Paen- - - - - - - - - - - - - - - --

Interval Not(Months) usual Intra- Intra-mural mural

1 89% 83% 67%12 87% 81% 64%24 87% 80% 63%36 87% 80% 63%60 87% . 80% 62%.6 12 18 24 30 36 42 48 54 60 66

Interval (Months) after Repair

FiGURE 3. Graph showing risk-adjusted time-related sur-vival after the arterial switch operation for simple transposi-tion of the great arteries (TGA) or TGA and ventricularseptal defect (VSD) according to the coronary arterialpattern(isobars). The depiction is a nomogram of a specific solutionof the multivariable equation (Table 4); "6 days" is the valueentered for age at repair, "no" for multiplicity of VSDs, and"no" for coexisting noncardiac anomalies.

n .v

---------------------------------

....... ..:...:. ...... :.

.....d .............. ............

1505

LA




- - -^

...................... ............. ....................... .............................

MuidpleInterval VSDs(months)

No yes1 89% 77%

1 2 88% 74%24 87% 74%36 87% 73%60 87% 73%

TABLE 5. Incremental Risk Factors for Death After ArterialSwitch Operation for Simple TGA and TGA With VSD*

Single hazardIncremental risk factors for death phase pPatient

LCA, LAD, or Cx arising from sinus 2Without an intramural course of LCA or LADt 0.15With an intramural course of LCA or LADt 0.03

Multiple VSDs 0.01

Coexisting noncardiac anomalies 0.04PA banding > 1 month previously 0.9Older age at repairt 0.5

Simple TGA. 0.07

SupportLonger global myocardial ischemic time 0.0005

ProcedureAorta transected distally 0.0006PA transected proximally or in midportion 0.09No Lecompte maneuver 0.001

Coronary implantation not at transection site 0.08


*Patient-specific, support, and procedural variables were en-tered into the nonparsimonious analysis. Risk factors identified inthe previous analyses in this sequential series were forced toremain in the equation, even though the probability value was

P>0.1.tThese are mutually exclusive variables, each with its ownp and

coefficient, and are subsets of "LCA, LAD, or Cx arising fromsinus 2."

tThis variable was active only in patients without PA banding> 1 month previously.§Simple TGA, but not TGA with VSD, was a positive interac-

tion term with age at repair.

Procedural Risk Factors for DeathThese are listed in Table 5.

Institutions and Their Experience as Incremental RiskFactors for Death

High-risk (n=11) and low-risk (n=7) institutionswere defined in an earlier study,3 and the risk-unad-justed differences in institutional mortality have per-sisted (Table 2). The same institutions were high-riskinstitutions in the current study when a risk-adjustingprocess of multivariable analysis was used (see TablesD-2, D-3, and D-4). Increasing institutional experiencewas associated with improvement in results in bothgroups of institutions (Table 6 and "Appendix Q"Table G1). The number of arterial switch operationsperformed in the group during the CHSS study (Figure6) was better correlated with survival than was thelength of time during which the arterial switch opera-tion had been performed (Figure G1).Not only was there a high-risk and a low-risk group of

institutions, but within each group there were institu-tional differences in the percent survival both early inthe experience (the intercept) and in the rate of changein the percent survival as experience increased (theslope) (Table 3 and Table G2). One low-risk institution(institution F) had somewhat less good results as the

n- ....................

v -

100

90

-80

9.5 70h-

9^ eoagCD 50z;340:F

19..: 30.299

20

10

0 6 12 18 24 30 36 42 48 54 60Interval (Months) after Repair

FIGURE 4. Graph showing risk-adjusted time-related sur-

vival after the arterial switch repairfor simple transposition ofthe great arteries (TGA) or TGA and ventricular septal defect(VSD) according to whether or not there are multiple VSDs(isobars). The depiction is a nomogram of a specific solutionof the multivariable equation (Table 4); "no" indicates eithersimple transposition or transposition with a single VSD, "6days" was entered as the value for age at repair, "no" as thevalue for coexisting noncardiac anomalies, and "usual" forthe coronary artery pattern.

total circulatory arrest, then both variables applied asrisk factors (Table D-5). The nature of the effect of theglobal myocardial ischemic time was such that, in pa-tients in whom the Lecompte maneuver was used, the6-month survival with ischemic times of 30, 60, 90, and120 minutes was predicted to be 96%, 94%, 90%, and83%, respectively.

100

ge

ge SilmeTGA

TGA th94 VSD

92%92 10 9921Z 91%

7 87% 88%90 14 85% 86%

28 83% 84%.........

...........................

82

800 3 6 9 12 15 le 21 24 27 30

Age (days) at Arterial Switch RepairFIGURE 5. Graph showing risk-adjustedpercent survivalforat least 6 months after the arterial switch repair for simpletransposition of the great arteries (TGA) or TGA withventricular septal defect (VSD) according to the age at repair(continuous variable along the horizontal axis) and whetherthe diagnosis was simple TGA or TGA and a single VSD(isobars). The difference in survival ofsirnple TGA and TGAwith VSD is a result of the former's being an interaction termin the risk factor equation (see footnote 9 in Table 4). Thedepiction is a nomogram of a specific solution of the multi~variable equation (Table 4); "6 days" was entered as the valueforageatrepair, "usual"forthecoronaryarterypattern, "no"for multiplicity of VSDs, and "no" for coexisting noncardiacanomalies.




TABLE 6. Incremental Risk Factors for Death After the ArterialSwitch Operation for Simple TGA and TGA with VSD*



Low-risk institutions 0.09High-risk institutions 0.0007

Intramural course of LCA or LADt 0.002Coexisting cardiac or noncardiac anomalies 0.06PA banding >1 month previouslyt 0.002Older age at repair§ 0.06

InstitutionsNumber of cases since first arterial switch repairll

Low-risk institutions 0.0001High-risk institutions <0.0001


*Only patient and institutional (high-risk vs. low-risk and expe-rience) variables were entered into the analysis for this parsimo-niously derived equation (n=513). The experience was in terms ofnumber of arterial switch operations performed on CongenitalHeart Surgeons Society patients.

tThis variable was active only in high-risk institutions.*This variable was active only in low-risk institutions.§Applies only to patients with simple TGA who have not

undergone PA banding > 1 month previously.iI"Number of cases" refers to the number of cases in the

individual patient's institution.

number of cases performed increased (Figure 7). Fourhigh-risk institutions had experience curves differentfrom those of the rest of the high-risk institutions. Twoof these (institutions N and R) had declining survivalwith increasing experience (Figure 8); one (institutionK) improved with increasing experience but at a lesserrate than did the remainder of the high-risk group (seeFigure 7); and one (institution Q) had a more rapid rateof improvement than did others in the high-risk group(see Figure 8). After 30 operations had been performed,the high-risk institutions, other than these four, hadpredicted survival rates similar to those of the low-riskinstitutions (see Figure 7). The findings were generallysimilar but different in a few details when duration ofexperience (rather than number of cases) was used asthe marker (Figures C-1-C-4).

Functional ClassAll surviving patients but one have a functional status

analogous to that described by New York Heart Asso-ciation functional class I; that is to say, all but one havethe physical strength and activity expected of normalinfants and children of their age. One child, 27 monthsold, tires clearly sooner than her peers at play anddevelops cyanotic lips.

DiscussionCritique of the StudyThe patient population is unique, incorporating as it

does a large number of patients entered within 2 weeksof birth, most of whom were less than 48 hours old atentry. Thus, the group comes as close as is possible to

100

90R o> 80

u= 70

LA

X 20

10

r~~~~~~~~~~~~~~~~~~~~-. - -- --/,<' ~~No.ofCHSS Xonk H -ik

.',/ ~~~Previously nitio Istuost/, ~~~~~~~063% 30%tIt ~ ~~~~~~~S 9W% 69%

10 95% 78%20 95% 85 %50 95% 91%100 95% (94%)

0 5 10 15 20 25 30 35 40Institutional Experience (number of previous

operations) with Arterial Switch Repair

45 50

FIGURE 6. Graph showing risk-adjusted predicted survivalfor at least 5years after the arterial switch operation for simpletransposition of the great arteries (TGA) or TGA withventricular septal defect (VSD) according to whether theoperation was performed in a low-risk or high-risk institution(isobars) and to the number of arterial switch operations inCongenital Heart Surgeons Society (CHSS) patients per-formed by the institution (a continuous variable along thehorizontal axis). The depiction is a nomogram of a specificsolution ofthe multivariable equation (Table 6); "6 days" wasentered as the value for age at operation, "usual" for thecoronary artery pattern, "no" for multiplicity of VSDs, and"no" for coexisting noncardiac anomalies.

being an unselected sample of newborns with simpleTGA. A few patients with TGA and VSD may not bediagnosed or may not seem to be sufficiently ill to bereferred to a pediatric cardiac surgical center within 2weeks of birth, but the patients in the study are probablya reasonably representative sample, even in this group.The study is an observational one, and treatment was

not randomly assigned. However, it encompasses aunique time period, early in which institutions were justbeginning to have an experience with the arterial switchoperation and at the end of which many had accumu-lated considerable experience with the procedure. Thisoptimized the possibility of finding patient-specific andprocedural incremental risk factors for the arterialswitch operation and allowed observations of thechange and rate of change in outcomes in individualinstitutions as experience grew. Although not a part ofthis analysis, the atrial switch operation was performedfrequently on patients in this study in the early years;future risk-adjusted comparisons can be made of out-comes with these two types of operations.The data set and the annual cross-sectional follow-

ups are considered to be as complete as is possible. Asmuch of the data is presented as is possible in an articleof reasonable length, but not all data could be included.However, multivariable analyses that are thoughtfullyperformed and carefully studied conserve the maximalamount of information and allow generation of the mostreliable inferences. Hazard function regression analysiswas used for most of these.8 Multiple internal valida-tions and comparisons of predicted and actual outcomesfor the hazard function regression analysis method havebeen published,10-'2 and one internal validation is pre-sented in this article (see Figure 2).

c) .




100

90

80

70

60

50

40

30

20

10

0

0S102050100

A100

90

R o

4n 70

20

80'# T0

~e 50

g 40

.2 20

10

59% 8% 59% 33%94% 54% 94% 76%95% 68% 95% 84%95% 79% 96% 90%93% (89%) 96% (95%)(86%) (93%) 96% (97%)

0 10 20 30 40 50 60 70 80

Institutional Experience (number of previousoperations) with Arterial Switch Repair

090 100

0 10 20 30 40 50 60 70 80 90 100


FIGURE 7. Graphs showing risk-adjusted predicted survivalfor at least 5 years after the arterial switch operation inpatients with simple transposition of the great arteries (TGA)or TGA with ventricular septal defect (VSD) according tocertain individual institutions (F and K) and the otherlow-risk and high-risk groups of institutions (isobars) and thenumber ofpreviously performed arterial switch operations in

Congenital Heart Surgeons Society (CHSS) patients (contin-uous variable along the horizontal axis). The depiction is a

nomogram ofa specific solution ofthe multivariable equation(Table 3); "6 days" was entered as the value for age atoperation, "usual" for the coronary artery pattern, "no" formultiplicity of VSDs, and "no" for coexisting noncardiacanomalies. Panel A: Annotated plot. Panel B: Labeled butotherwise unannotatedplot in which institution F is omitted toshow that the confidence intervals of "other high-risk institu-tions" overlap with those of "other low-risk institutions" after30 operations per institution.

The important outcome event "reintervention" hasalso been studied, but to avoid overloading the infor-mation content of the present article, this will bepresented later.

Applicability of the Arterial Switch OperationThe study indicates that the arterial switch operation

can be performed particularly safely in the first week oflife (see Figure 5), which is important because of thelethality of the malformations, particularly in patientswith simple TGA, among whom 5-10% are dead within30 days of birth when untreated surgically and in spite of

,,,, , , ,:~~~~~~..:......:.: ...... . ..:...:'.:.:.:.:.:.:..':...

17, , /No. o f Othe~~~~~~rOthere7 t , / Cases /N Q /~~~~Rlow- high-

,''/, - - Previously// rikis/,< ~ ~~~~~O20% 33 52%9% 33%>'#'_ = ~~~5/10% (99.7)19 94% 76%>/~~~~~~t 0 /(0%) (10% 1) 95% 84%t ,- \ ~~ (20) (0%) (100)1(% (96%) (9O%)~~~~~~(50)/ (0%) (1 0 0){%(96%) (95%)> ... x (~~1400 (0%) (1 0 0t{%(96%) (97%)

....................

2 3 4 5 6 7 8Institutional Experience (number of previous

operations) with Arterial Switch Repair

9 10

2 3 4 5 6 7 8


FIGURE 8. Plots similar to those in Figure 7, representing theother institutions appearing in the multivariable equationdepicted in Table 3. Panel A: Annotated plot. Panel B:Unannotated plot to demonstrate the rightward extension ofthe lines. CHSS, Congenital Heart Surgeons Society.

balloon atrial septostomy.,2 The operation can be per-formed in the presence of any coronary artery branch-ing pattern, as was demonstrated earlier by Quaegebeurand colleagues,13 and even when the aorta is posterior.6However, a few institutions elected not to perform thearterial switch operation in the presence of certaincoronary patterns.

Early and Intermediate-Term OutcomesThe early and intermediate-term outcomes of the

arterial switch operation for simple TGA and TGA withVSD can be outstandingly good. A decade ago it wouldhave been difficult to believe that 97% of patients bornwith simple TGA or TGA with VSD and the usualcoronary anatomy could be corrected, alive, and pre-dicted to be well 5 years after the operation (Figure 9).

Risk Factors for Poor OutcomesIt is remarkable that there are so few patient-specific

risk factors for premature death. Multiplicity of coexistingVSDs probably continues to be a risk factor. An intramu-ral left coronary artery has been a strong risk factor fordeath (see Figure 3), and this has been found by others aswell.14'15 This is an uncommon variant (see Table 1), andin the current era even this can probably be bettermanaged than was the case earlier.14 Other unusual

A

.E3:

0

am

10

B100

90

'> 70L.S0!60

a 50

~40

*2 20

10

j -ffiJ - - -n 9 AfEXv

| l X | . - l s z l * *l~~~~~~~A

%i




Interval Percent(Years)

1112 98%1 97%3 97%5 97%

(10 97%)

outcomes by the identification, at the earliest possibletime, of changes in risk-adjusted outcomes across time.Preliminary analyses of the experiences described inthis study indicate that changes across time could havebeen identified with reasonable certainty after an insti-tution had performed 10 arterial switch operations(E.H. Blackstone and J.W. Kirklin, unpublished obser-vations, 1991).

Appendix APrevalences of Patient Characteristics

1 2 3 4 5 6 7 8 9Interval (Years) After ARTERIAL SWITCH REPAIR

10

FIGURE 9. Graph showing risk-adjusted time-related per-cent survival after an arterial switch repair in a neonate withsimple transposition of the great arteries (TGA) or TGA withventricular septal defect (VSD) operated on in an institutionof proven competence with the arterial switch repair. Thedepiction is a nomogram of a specific solution of the multi-variable equation described in Table C-i; "3.4 kg" was thevalue entered for birth weight, "usual" for coronary arterypattern, and "no" for coexisting noncardiac anomalies.

coronary patterns were weak risk factors throughout thisexperience (see Figure 3).

In an analysis specifically designed to separately testthe components of the support systems (see Table D-5),the duration of total circulatory arrest was found to bea risk-adjusted risk factor for death. This supports aplan of performing the arterial switch operation withlittle or no use of profoundly total circulatory arrest.13,'5The incremental risk of longer duration of global myo-cardial ischemia suggests that improved techniques ofmyocardial management may improve results overall.The procedural risk factors identified in this study

endorse the usual practice of many surgeons of transect-ing the aorta in its midportion and the pulmonary trunkas far distally as is possible, using the Lecompte maneu-ver, and implanting the coronary arteries into theneoaorta at or near the transection site. Failure to do sodecreased the risk-adjusted survival (see Table D-1).

Institutions and Their ExperienceOutcomes after the arterial switch operation have

varied, in this study, according to the institution inwhich it was performed and according to the institu-tion's experience. This was the result of some variabilityin outcomes in the institutions' initial experiences andan even greater variability in the rate of institutionalimprovement with further experience. As a result, someinstitutions that were high-risk early in the study rapidlybecame low-risk institutions, whereas others achievedimprovement more slowly; still others did not improveat all or worsened. In at least one initially low-riskinstitution, the results appear to be slightly less goodwith increasing experience. These differences betweeninstitutions appear to result in part from differinginstitutional responses to initially disappointing results.Probably such differences are not unique to the currentstudy.

Speculatively, physicians, institutions, and health careproviders could improve their responses to unfavorable

100lC.5g

' 90

R 80

: 70

OK 60

550

~340b. 300

20

CL

Age (days) at Entry

FIGURE A-1. Graph ofcumulativefrequency distribution ofthe age at entry ofpatients with simple transposition of thegreat arteries (TGA) and TGA with ventricular septal defect(VSD) who underwent an arterial switch operation.

100

90

t:70

~~~Birth ight50Percentile TGA

~~~~~TGAVSD

30 ,{ 10 2.7 2.8e' 20 84

20 so 3.4 3.5~~~~~753.8 3.7

io ~~~90 4.0 4.2

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5

Birthweight (kg)

FIGURE A-2. Graph showing cumulative frequency distri-bution of the birth weight ofpatients with simple transpositionof the great arteries (TGA) and TGA with ventricular septaldefect (VSD) undergoing an arterial switch operation.

100

.g90a=80Q^

' 70

eU

0° 40CL

20

= 10

00

:::::: ............... M!!




100

ap 90

*0 80

70c

0.C eo

40

0

X 30

20

C. 10

Age ~TGA(days) Simple with

TGA VSD

2 4.2% 0.8%4 25% 14%7 63% 36%14 94% 58%30 97% 67%

3 6 9 12 15 18 21

Age (Months) at Arterial Switch24 27 30

FIGURE A-3. Graph showing cumulative frequency distri-bution ofthe age at the time ofthe arterial switch repair. TGA,transposition of the great arteries; VSD, ventricular septaldefect.

TABLE A-1. Associated Cardiac Malformations in PatientsUndergoing the Arterial Switch Operation for Simple TGAand TGA With VSD

Total deaths

Confidenceinterval

Cardiac malformation* n No. % (%)

Coarctation 22 6 27 17-40

Interrupted aortic arch 1 0 0 0-85Hypoplastic transverse arch 8 1 12 2-36Multiplicity of VSD 19 7 37 24-51AV valve anomalies

Tricuspid 9 3 33 15-56Mitral 3 2 67 24-96

Pulmonary valve and LVOT anomalies 13 2 15 5-33Aortic valve and RVOT anomalies 5 1 20 3-53Hypoplastic right ventricle 5 2 40 14-71Systemic venous anomalies 11 3 27 12-47Juxtaposition (left) of atrial appendages 9 3 33 15-56Pulmonary venous stenosis 1 0 0 0-85RCA to RA fistula 1 0 0 0-85Positional cardiac anomalies 7 3 43 20-68

TGA, transposition of the great arteries; VSD, ventricularseptal defect; AV, atrioventricular; LVOT, left ventricular outflowtract; RVOT, right ventricular outflow tract; RCA, right coronaryartery; RA, right atrial.

*Categories are not mutually exclusive.

Appendix BVariables Entered Into the Multivariable Analysis

All variables are dichotomous ones unless the units aregiven in parentheses; not all variables were entered into everyanalysis.

Patient VariablesDemography. Sex, birth weight (kilograms), age (days) at

arterial switch repair.Cardiac morphology. Type ofTGA (simple versus TGA with

VSD), multiplicity of VSDs, right ventricular outflow obstruc-tion (defined as coexisting subaortic stenosis and/or coarcta-

TABLE A-2. Associated Noncardiac Malformations in PatientsUndergoing the Arterial Switch Operation for Simple TGA andTGA With VSD

Total deaths

Confidenceinterval

Noncardiac malformation* n No. % (%)Persistent fetal circulation 4 3 75 37-97Tracheoesophageal fistula 1 0 0 0-85Multiple cervical and thoracic bonyanomalies 1 1 100 15-100Facial and cervical anomalies 1 1 100 15-100Microcephaly, choanal atresia 1 0 0 0-85Congenital adrenal insufficiency 1 0 0 0-85Bilateral club foot 1 0 0 0-85Goldenhar's syndrome 1 1 100 15-100

TGA, transposition of the great arteries; VSD, ventricularseptal defect.*The categories are not mutually exclusive.

TABLE A-3. Position of the Great Arteries in PatientsUndergoing the Arterial Switch Operation for SimpleTGA and TGA With VSD

Position of great arteries No. Percent of 330Ao anterior 00 203 62Ao anterior 30°R 54 16Ao anterior 60°R 41 12Ao-PA side-by-side 90°R 24 7

Ao anterior 30°L 6 2Ao anterior 60°L 2 0.6

Subtotal 330 100

Unknown 183

Total 513

TGA, transposition of the great arteries; VSD, ventricularseptal defect; Ao, aorta; PA, pulmonary artery; R, right; L, left.

tion and/or interrupted arch), coarctation, interrupted aorticarch, aorta directly anterior to pulmonary artery (PA), aortaanterior and to the right of PA, aorta and PA side by side (bothas dichotomous and ordinal variables), PA size divided byaortic size (mm), malalignment of aortic and pulmonic com-missures, positional cardiac anomalies, atrioventricular valveanomalies, hypoplastic right ventricle, systemic venous anom-alies, important coexisting noncoronary cardiac anomalies (asa group).Important coexisting noncardiac anomalies. Unspecified.Coronary artery pattern. (1 LADCx; 2R), (1-; 2LADCxR),

(1RLAD; 2Cx), (1LAD; 2CxR), LAD or Cx arising from sinus2, intramural course, coronary arteries arising from a singlesinus, number of coronary ostia (number), eccentric coronaryostium. Abbreviations are as follows: LAD, left anteriordescending coronary artery; Cx, circunmflex coronary artery;R, right.

Previous procedures. Atrial septostomy, banding of pulmo-nary trunk with delay > 1 month to arterial switch, banding ofpulmonary trunk with delay <1 month to arterial switch.

n - --- . .-.v~



Kirklin et al Risk Factors in Arterial Switch Operation for Transposition

Institutional VariablesHigh-risk institution for arterial switch operation (the method

of identification of high-risk institutions and low-risk institutionsis given in a previous publication3), each individual institution,number (number) of previous arterial switch operations withinCHSS at each institution, interval (months) since first CHSSarterial switch repair at each institution, date of entry ofpatients into study (months since January 1, 1985), date ofarterial switch operation (months since January 1, 1985).

Support VariablesUse of total circulatory arrest (TCA), TCA >30 minutes,

duration of TCA (minutes), lowest temperature ('C) duringrepair, global myocardial ischemic time (minutes). (Note thatglobal myocardial ischemic time in most analyses applied to allpatients, including those in whom a period of total circulatoryarrest was used. In one analysis [Table D-5], this variable wasapplied only to patients with no total circulatory arrest.)

Procedure VariablesLevel of aortic transection (proximal, mid, or distal, both as

dichotomous and ordinal variables), level of PA transection(proximal, mid, or distal, both as dichotomous and ordinalvariables), coronary explantation from transection site, coro-nary implantation adjacent to transection site, preparation forcoronary implantation (excision, incision, flap), sinus recon-struction (individual patches, single scalloped patch, no patchbut scalloped suture line), Lecompte maneuver, division of rightventricular coronary branch of left coronary artery.

Appendix CMultivariable Parsimoniously Derived Risk FactorEquations for Death, Incorporating Patient andInstitutional Variables and Duration of InstitutionalExperience With the Arterial Switch Repair inCongenital Heart Surgeons Society Cases Intothe Analysis

See Figures C-1-C-4 and Tables C-1 and C-2.

5> 60 _1_9% _9

o ~~~~~397% 83%

X ~~~~~597% 93%

, 10

00o 05s 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0Institutional Experience (year since firstoperation) with Arterial Switch Repair

FIGURE C 1. Graph showing risk-adjusted predicted sur-vival for at least 5 years after the arterial switch operation forsimple transposition of the great arteries (TGA) or TGA withventricular septal defect according to the duration of theinstitutional experience with the arterial switch repair withinthe Congenital Heart Surgeons Society study (a continuousvariable along the horizontal axis) and the category of theinstitution (isobars). The depiction is a nomogram of aspecific solution of the multivariabke equation (Table C-i);"3.4 kg"' was entered for birth weight, "usual"'for the coronaryartery pattern, and "no"for coexcisting noncardiac anomalies.

100

90

.2 80

iL 70

. 60

U40

10

- ,-: - Insti tion,, - /,- _ _ _- Interval orOth er(years) 0 High Low, Risk Risk

0 55% 55% 66%0.5 53% 67% 97%1 52% 77% 97%2 50% 89% 97%3 47% 95% 97%_ -- - 4 (45%) 98% 97%5 (42%) 99% 97%

o ..................a............... .0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Institutional Experience (years since firstoperation) with Arterial Switch Repair

FIGURE C-2. Graph showing risk-adjusted predicted sur-vival for at least 5 years after the arterial switch operationaccording to certain individual institutions and the interval(years) since thefirst arterial switch operation on a CongenitalHeart Surgeons Society patient in that institution. The depic-tion is a nomogram of a specific solution of the multivariableequation (Table C-2) in which the values enteredfor the othervariables were the same as for Figure C-i.

100l..90

2 80

70

gm 60LAzc 50

a 40

F 30

.2 20

10

,,--' tit~~~~~ution,, / lf~~~~ytearrsa) g>> ther

, S ' /, # ' / Rlsis~~~Risk/' ~~~~~~~0/22% 55% 66%i,S ~~~~~~~0.5/23# 67% 97%,"^ ~~~~~~1/24% 77% 97%

' '^_ ~~~~~~~2/27% 89% 97%-___ _ _ _ 3 /30% 95% 97%

4 33%) 98% 97%5J(36%) 99% 97%

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0Institutional Experience (years since firstoperation) with Arterial Switch Repair

FIGURE C-3. Graph showing risk-adjusted predicted sur-vivalfor at least 5 years after the arterial switch operation; thedepiction is similar to Figure C-2 but with a different specificinstitution.

1 vO F-

90

.k 80

4^ 7070> 60

at 50

cm~ 40

~'O30.v 20

10

(years) \K °?heNte,^ o/ , ~~~~~~~~~RiskRisk

' ,'/ 8 ~ ~~~016% ss% 66%/ /,' ~~~~0.5 29% 67% 97%w / 8 ~~~~~144% 77% 97%

2 69% 89% 97%3 85% 95% 97%

4 93% 98% 97%5 97% 99% 97%

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5Institutional Experience (years since firstoperation) with Arterial Switch Repair

5.0

FIGURE C-4. Graph showing risk-adjusted predicted sur-vivalfor at least 5 years after the arterial switch operation; thedepiction is similar to Figure C-2 but with a different specificinstitution.

o ...........................a....a.... ......... ....a.... A.... a........................ ._,

1----`

liw,-

n P --..... ... ....

1511




TABLE C-1. Incremental Risk Factors for Death After theArterial Switch Operation for Simple TGA and TGA With VSD*


PatientLCA, LAD, or Cx arising from sinus 2tIntramural course of LCA or LAD:Coexisting cardiac anomalies (includingmultiple VSDs)Coexisting noncardiac anomalies

PA banding > 1 month previously§Lower birth weighti

InstitutionsLesser interval (years) since firstarterial switch repairLow-risk institutionsHigh-risk institutions

0.00070.07

0.070.070.0090.05

<0.00010.0004


*This equation is entirely analogous to that in Table 6, exceptthat in this equation the institutional experience was in terms ofthe interval since the first arterial switch operation for a Congen-ital Heart Surgeons Society patient in the individual patient'sinstitution.

tThis variable applies to those without or with an intramuralcourse.

*This variable was active only in high-risk institutions and addedan increment of risk to "origin from sinus 2."

§This variable was active only in low-risk institutions.liThis variable was active only in simple TGA in low-risk

institutions.

TABLE C-2. Incremental Risk Factors for Death After theArterial Switch Operation for Simple TGA and TGA With VSD$


PatientLCA, LAD, or Cx arising from sinus 2 0.0004Intramural course of LCA or LADt 0.02Coexisting cardiac anomalies(including multiple VSDs) 0.004Coexisting noncardiac anomalies 0.096PA banding > 1 month previously? 0.02Lower birth weight§ 0.09

InstitutionsLesser interval (years) since firstarterial switch repair

Low-risk institutionsll <0.0001High-risk institutions¶ <0.0001

Institution K# 0.003Institution M# 0.002Institution N* * 0.002Institution 0** 0.0004Institution P# 0.009Institution R** <0.0001


*This equation is entirely analogous to that in Table 3, exceptthat in this equation the institutional experience was in terms ofthe interval since the first arterial switch operation for a Congen-ital Heart Surgeons Society patient.

tThis variable was active only in high-risk institutions. Its effectwas additive to that of the "origin of the artery from sinus 2."

tThis variable was active only in low-risk institutions.§This variable pertained only to simple TGA.IlNo institutional differences (interactions) were found in this

group.¶The following institutions are interaction terms with "interval

in high-risk institutions." The "interval" is as defined in Table C-1.#The interaction affected only the intercept, and the coefficient

was positive (meaning that the prevalence of death at "time 0" wasgreater and the survivorship less).

**The interaction affected only the slope, and the coefficientwas positive (meaning that the prevalence of death was greateralong the slope and the survivorship less).




Appendix DSequentially and Nonparsimoniously DerivedMultivariable Risk Factor Equations for Death

See Tables D-1-D-5.

TABLE D-1. Incremental Risk Factors for Death After ArterialSwitch Operation for Simple TGA and TGA With VSD*


PatientLCA, LAD, or Cx arising from sinus 2Without an intramural course of LCA or LAD 0.099

With an intramural course of LCA or LAD 0.02

Multiple VSDs 0.01

Coexisting noncardiac anomalies 0.09PA banding > 1 month previously 0.7

Older age at repairt 0.11

Simple TGAf 0.21

ProcedureAorta transected distally 0.002PA transected proximally or midportion 0.06No Lecompte maneuver 0.0003Coronary implantation not at transection site 0.05


*Patient-specific and procedural variables were entered into thenonparsimonious analysis. Risk factors identified in the previousanalyses in this sequential series were forced to remain in theequation, even though the probability value was p>0.1.

tThis variable was active only in patients without PA banding> 1 month previously.tAn increment in risk was added when the patient had simple

TGA rather than TGA with VSD.



PatientLCA, LAD, or Cx arising from sinus 2Without an intramural course of LCA or LAD 0.002With an intramural course of LCA or LAD 0.07


Simple TGAf 0.7Institutions (positive coefficients)§H 0.004I 0.008J 0.002K <0.0001L 0.04M <0.0001N <0.00010 0.0007p <0.0001Q 0.01R <0.0001


*Patient-specific and institutional variables were entered intothe nonparsimonious analysis. Risk factors identified in the previ-ous analyses in this sequential series were forced to remain in theequation, even though the probability value wasp>0.1.

tThis variable was active only in patients without PA banding>1 month previously.*An increment in risk was added when the patient had simple

TGA rather than TGA with VSD.§These turn out to be the same institutions as those originally

defined as high risk in a non-risk-adjusted manner.3






PatientLCA, LAD, or Cx arising from sinus 2Without an intramural course of LCA or LAD 0.001With an intramural course of LCA or LAD 0.008

Multiple VSDs 0.004Coexisting noncardiac anomalies 0.03PA banding >1 month previously 0.10Older age at repairt 0.9

Simple TGAt 0.3SupportLonger myocardial ischemic time 0.01

Institutions§H 0.05I 0.13J <0.0001K <0.0001L 0.04M <0.0001N <0.00010 0.0002P 0.0006Q 0.001R <0.0001

Institutional experience (number of cases)IIE 0.04F 0.098J 0.05K 0.0003Q 0.06


*Patient-specific, support, and institutional variables were en-tered into the nonparsimonious analysis. Risk factors identified inthe previous analyses in this sequential series were forced toremain in the equation, even though the probability value wasp>0.1.


TGA rather than TGA with VSD.§In this analysis, no prior grouping into high-risk and low-risk

institutions was made. The institutions not listed could each beconsidered to be part of a group called "all others." The coeffi-cients are all positive (increased risk of death).

liThese are interaction terms between the individual institutionand its number of cases, even though the institution is in the "allothers" group. "Number of cases" is as defined in Table 6.


Incremental risk factors for death

PatientLCA, LAD, or Cx arising from sinus 2Without an intramural course of LCA or LADWith an intramural course of LCA or LAD

Multiple VSDsCoexisting noncardiac anomaliesPA banding >1 month previouslyOlder age at repairt

Simple TGA*SupportLonger myocardial ischemic time

ProcedureAorta transected distallyPA transected proximally or midportionNo Lecompte maneuverCoronary implantation not at transection siteSinus reconstruction with individual patches§

Institutionsil

HIJKLMN0p

Single hazardphase p

0.0090.070.0010.140.40.70.5

0.005

0.010.210.0020.80.009

0.060.040.90.010.0020.040.0004

<0.00010.0070.001

Q 0.14R <0.0001


*Patient-specffic, support, procedural, and institutional vari-ables were entered into the nonparsimonious analysis. Risk factorsidentified in the previous analyses in this sequential series wereforced to remain in the equation, even though the probabilityvalue wasp>0.1.


TGA rather than TGA with VSD.§This variable was significant for the first time in this analysis.IlInstitutional experience was not entered into this analysis.lMITe coefficient had a negative sign, indicating a lesser overall

prevalence of death. All other institutions had a positive coefficient.

S







Without an intramural course of LCA or LAD 0.07With an intramural course of LCA or LAD 0.04


Simple TGA* 0.06SupportLonger total circulatory arrest time 0.03Longer myocardial ischemic time§ 0.03


*Patient-specific and support variables only were entered intothe sequential nonparsimonious series of analyses.

tThis variable was active only in patients without PA banding>1 month previously.tAn incremental risk was added when the patient had simple

TGA rather than TGA with VSD.§In this analysis the variable "myocardial ischemic time" ap-

plied only to patients in whom there was no total circulatory arrest.(In patients with a period of total circulatory arrest, myocardialischemic time was at least as long as the total circulatory arresttime.)

AcknowledgmentsThe authors express their appreciation of the work over a

5-year period of the pediatric cardiologists, cardiac surgeons,nurses, and coordinators in the institutions participating in thisstudy of transposition by the Congenital Heart SurgeonsSociety. The institutions, in random order, are Mott Children'sHospital at the University of Michigan Medical Center; Uni-versity of Alabama Medical Center; The Boston Children'sHospital; The Children's Hospital in Buffalo, N.Y.; Universityof Chicago and Michael Reese Hospital Medical Center;Children's Memorial Hospital in Chicago; Children's Hospitalof Michigan in Detroit; The Penn State College of Medicine(The Milton S. Hershey Medical Center); University of IowaHospital and Clinics; Children's Hospital of Los Angeles;Miami Children's Hospital; University of Miami (and JacksonMemorial Hospital) in Miami, Fla.; The Montreal Children'sHospital; Columbia-Presbyterian Medical Center in NewYork; Children's Hospital of Philadelphia; Oregon HealthSciences University; Mayo Clinic; University of California SanFrancisco; University of Utah Medical Center (and PrimaryChildren's Medical Center); All Children's Hospital in St.Petersburg, Fla.; Hospital for Sick Children in Toronto; Uni-versity of California Los Angeles; British Columbia Children's

Hospital in Vancouver. The authors are deeply indebted toMary Lynne Clark, Phyllis Newsom, and Gail Mertz for theirmanagement of the follow-up, to Rob Brown for the datamanagement and some of the analyses, and to Debbie Nubyfor her skill in developing the graphics and the manuscript.They appreciate also the help of Dr. Jan Quaegebeur inidentifying the coronary anatomy from the surgical reportsand in understanding the variations in the technique of thearterial switch operation that were used.

References1. Trusler GA, Castaneda AR, Rosenthal A, Blackstone EH, Kirklin

JW, and the Congenital Heart Surgeons Society: Current results ofmanagement in transposition of the great arteries, with specialemphasis on patients with associated ventricular septal defect.JAm Coll Cardiol 1987;10:1061-1071

2. Castaneda AR, Trusler GA, Paul MH, Blackstone EH, Kirklin JW,and the Congenital Heart Surgeons Society: The early results oftreatment of simple transposition in the current era. J ThoracCardiovasc Surg 1988;95:14-28

3. Norwood WI, Dobell AR, Freed MD, Kirklin JW, Blackstone EH,and the Congenital Heart Surgeons Society: Intermediate resultsof the arterial switch repair. J Thorac Cardiovasc Surg 1988;96:854-862

4. Gittenberger-de Groot AC, Sauer U, Oppenheimer-Dekker A,Quaegebeur J: Coronary arterial anatomy in transposition of thegreat arteries: A morphologic study. Pediatr Cardiol 1983;4(suppl1):15-24

5. Gittenberger-de Groot AC, Sauer U, Quaegebeur J: Aortic intra-mural coronary artery in three hearts with transposition of thegreat arteries. J Thorac Cardiovasc Surg 1986;91:566-571

6. Tam S, Murphy JD, Norwood WI: Transposition of the greatarteries with posterior aorta: Anatomic repair. J Thorac CardiovascSurg 1990;100:441-444

7. Kaplan EL, Meier P: Nonparametric estimation from incompleteobservations. JAm Stat Assoc 1958;53:457-481

8. Blackstone EH, Naftel DC, Turner ME Jr: The decomposition oftime-varying hazard into phases, each incorporating a separatestream of concomitant information. J Am Stat Assoc 1986;81:615-624

9. Ferrazzi P, McGiffin DC, Kirklin JW, Blackstone EH, Bourge RC:Have the results of mitral valve replacement improved? J ThoracCardiovasc Surg 1986;92:186-197

10. Kirklin JW, Blackstone EH, Bargeron LM Jr, Pacifico AD, KirklinJK: The repair of atrioventricular septal defects in infancy. Int JCardiol 1986;13:333-351

11. Crawford ES, Kirklin JW, Naftel DC, Svensson LG, Coselli JS, SafiHJ, Hess KR: Surgery for acute ascending aortic dissection:Should the arch be included? J Thorac Cardiovasc Surg 1992;104:46-59

12. Kirklin JW, Blackstone EH, Jonas RA, Shimazaki Y, Kirklin JK,Mayer JE, Pacifico AD, Castaneda AR: Morphologic and surgicaldeterminants of outcome events after repair of tetralogy of Fallotand pulmonary stenosis: A two-institution study. J Thorac Cardio-vasc Surg 1992;103:706-723

13. Quaegebeur JM, Rohmer J, Ottenkamp J, Tuis T, Kirklin JW,Blackstone EH, Brom AG: The arterial switch operation: An eight-year experience. J Thorac Cardiovasc Surg 1986;92:361-384

14. Day RW, Laks H, Drinkwater DC: The influence of coronaryanatomy on the neonatal arterial switch operation. J Thorac Car-diovasc Surg 1992;104:706-712

15. Lupinetti FM, Bove EL, Minich LL, Snider AR, Callow LB,Meliones JN, Crowley DC, Beekman RH, Serwer G, Dick M,Vermilion R, Rosenthal A: Intermediate-term survival and func-tional results after arterial repair for transposition of the greatarteries. J Thorac Cardiovasc Surg 1992;103:421-427



clinical outcomes after the arterial switch operation for

Documents