Br Heart J 1989;62:43-9

Colour flow imaging in the diagnosis of multipleventricular septal defectsGEORGE R SUTHERLAND, JOHN H SMYLLIE, BRUCE C OGILVIE,BARRY R KEETON

From the Department of Paediatric Cardiology, Wessex Cardiothoracic Unit, Southampton General Hospital,Southampton

SUMMARY Thirty one patients with multiple ventricular septal defects were studied by crosssectional echocardiography, conventional pulsed and continuous wave Doppler, colour flowimaging, and left ventriculography to determine the relative diagnostic benefits and pitfalls of eachtechnique. The patients studied had a wide range of congenital heart defects with 19 patientshaving isolated multiple ventricular septal defects, three with associated tetralogy of Fallot, fivewith double outlet right ventricle, three with complete transposition and ventricular septal defect,and one with a complete atrioventricular septal defect. In 23 patients the defects were inspected atoperation. Cross sectional imaging with integrated pulsed and continuous wave Doppler correctlyidentified multiple defects in only 12 (39%) patients. In contrast, colour flow imaging was accuratein 24 (77%) patients and left ventriculography in 20 (65%) patients. When patients weresubdivided on the basis of relative peak systolic ventricular pressures into restrictive defects (18patients) and non-restrictive defects (13 patients) the diagnostic value of colour flow imaging wasdifferent for each group. Colour flow mapping correctly identified multiple ventricular septaldefects in 16/18 (89%) patients with restrictive defects but only 8/13 (62%) with non-restrictivedefects. The comparative diagnostic accuracy of left ventriculography was 15/18 (83%) in therestrictive group and 5/13 (38%) in the non-restrictive group.

Colour flow imaging was the single investigative technique with the greatest diagnostic accuracyin the diagnosis ofmultiple ventricular septal defects. It failed to be consistently accurate in definedsubgroups with non-restrictive defects as did left ventriculography. The greatest overall diagnosticaccuracy in this series was obtained when both colour flow imaging and ventriculographytechniques were used in combination in a complementary fashion.

Multiple ventricular septal defects are not un-common in congenital heart disease. The reportedfrequency in patients with multiple ventricular septaldefects as an isolated lesion ranges from 4% to18%.' Multiple ventricular septal defects are morecommon in patients with double outlet right ven-tricle,4 complete atrioventricular septal defects,56tetralogy of Fallot,' and complete transposition withventricular septal defect.5

It is important to make a precise preoperativediagnosis of multiple ventricular septal defects

Requests for reprints to Dr George R Sutherland, Thoraxcentre,Dijktzigt Academic Hospital, 3000 DR Rotterdam, TheNetherlands.

Accepted for publication 24 January 1989

because they are associated with a higher surgicalmortality23 and a higher rate of palliative pulmonaryartery banding rather than primary repair; failure tomake an accurate preoperative diagnosis leads to ahigh reoperation rate.7 Traditionally the diagnosis ofmultiple ventricular septal defects has relied on leftventriculography.' Despite recent improvements inthis technique such as the introduction of biplaneequipment for axially angled projections and thedecreased toxicity of contrast agents (allowing mul-tiple injections), the diagnostic accuracy of leftventriculography at best only approaches 86%.'

Ultrasound studies have shown that cross sectionalechocardiography can visualise most moderate tolarge defects especially when they are sited in theperimembranous and the smooth muscular inletsepta.9 However, small defects both single and mul-

43

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from

44

tiple within the trabecular and muscular outlet septahave proved difficult to visualise by cross sectionalimaging alone.9 The addition of pulsed Doppler toultrasound systems has improved the diagnosis ofsmall defects by identifying the diagnostic disturb-ances of transseptal and right ventricular flowassociated with these defects.10 Although it has beensuggested that multiple defects can be detected by a

right ventricular and septal scanning technique withpulsed Doppler,'0 others have shown that such an

integrated technique of cross sectional imaging andpulsed Doppler scanning can be inaccurate in thediagnosis of multiple ventricular septal defects."Both pulsed and continuous wave Doppler tech-niques are inadequate where non-restrictive ven-tricular septal defects are present. These result in lowtransseptal flow velocities and a laminary flow acrossthe defect,'0 in contrast with the high velocity tur-bulent flow associated with restrictive defects.

Colour flow imaging was a useful technique for thediagnosis of single restrictive ventricular septaldefects because it allowed visualisation of theircharacteristically turbulent transseptal and rightventricular flow patterns.'2 Previous work usingcolour flow imaging in the diagnosis of multipleventricular septal defects (mainly confined to restric-tive defects) suggested both a high sensitivity andspecificity for their detection." However, no directcomparison between the predictive value of colourflow imaging and left ventriculography could bemade in the reported series, because only patientswith multiple ventricular septal defects previouslyidentified by left ventriculography were studied.Also no comparison was made between the diagnosticrole of colour flow imaging in groups with restrictiveand non-restrictive defects.The aims of the present study were to evaluate the

usefulness of colour flow imaging in the diagnosis ofall types of multiple ventricular septal defects, tocompare this prospectively with the diagnosticaccuracy of left ventriculography, and to define anyinherent limitations in the technique.

'Patients and methods

STUDY PATIENTS AND TYPE OF VENTRICULAR

SEPTAL DEFECTBetween June 1986 and October 1987, 31 infants andchildren (aged 2 days to 11 years, mean 2- 11 years)with multiple ventricular septal defects diagnosed atthe Wessex Paediatric Cardiac Unit were inves-tigated by angiography, cross sectional echocar-diography, and colour flow imaging. Nineteenpatients had isolated multiple ventricular septaldefects, 11 ofwhom had a combination of perimem-branous and muscular defects, six had multiple

Sutherland, Smyllie, Ogilvie, Keetondiscrete muscular defects, and two had "Swisscheese" defects. Of the 12 with ventricular septaldefects occurring as part of a complex congenitallesion included in the series, three patients withtetralogy of Fallot and three patients with doubleoutlet right ventricle had a combination of a peri-membranous outlet plus muscular ventricular septaldefects. Two further patients with double outletright ventricle had non-committed multiple mus-cular defects. Three patients with complete trans-position and multiple ventricular septal defects had acombination of a single perimembranous plus amuscular defect and the patient with a completeatrioventricular septal defect had an associated mus-cular defect (table 1).

Patients were entered into the study only if anunequivocal diagnosis of multiple ventricular septaldefects was first made by: (a) an unequivocallydiagnostic left ventriculogram (16 patients), (b) Crosssectional imaging plus positive colour flow imaging(10 patients), and (c) surgical inspection (five patientshad multiple ventricular septal defects identified forthe first time at operation). All patients then under-went any ofthe other preoperative investigations notalready undertaken, so that every patient in the studyhad available for review cross sectional echocardio-graphy with conventional Doppler studies, colourflow imaging studies, and left ventriculography. Thefive patients whose diagnosis of multiple ventricularseptal defects was first made at operation all hadpreoperative left ventriculograms, cross sectionalimaging, and conventional Doppler and colour flowstudies, which were then reviewed. To date a total of23 patients have had surgical closure of their ven-

Table 1 Morphological diagnosis versus type of ventricularseptal defect

Restrictive Non-restrictiveMorphologicaldiagnosis Defect type No Defect type No Total

Isolated VSDs PMO + musc 11(no other lesion) Mult musc 5 Mult musc 1 19

"Swiss cheese" 2Tetralogy of

Fallot 0 PMO + musc 3 3DORV 0 PMO + musc 3

Mult musc 2 5TGA/VSD PMO + musc 1 PMO + musc 1

PMI + musc 1 3CAVSD 0 PMI + musc 1 1Total 18 13 31

VSD, ventricular septal defect; PMO + musc, perimembranousoutlet plus muscular defects; Mult musc, multiple musculardefects; PMI + musc, perimembranous inlet plus musculardefects; DORV, double outlet right ventricle; TGA/VSD,complete transposition of the great arteries with associatedventricular septal defect; CAVSD, complete atrioventricular septaldefect.

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from

Colourflow imaging in the diagnosis of multiple ventricular septal defects

tricular septal defects with concomitant inspection ofthe defect, and a further five patients have hadpulmonary artery banding as an initial procedure.

METHODSEchocardiographic studies were performed in thestandard manner as described in our work on theimaging of isolated ventricular septal defects.9 Crosssectional images were obtained with an ATLUltramark IV mechanical sector scanner with 5 0 and7-5 MHz transducers, to achieve high resolutionimages. Conventional Doppler (both pulsed andcontinuous wave) and colour flow imaging studieswere then obtained by a Toshiba SSH 65 A with 2-5,3-7, and 5 0 MHz phased array transducers. Colourflow studies were performed in both the power andthe variance modes with appropriate gain and filtersettings. The diagnosis by colour flow imaging of asingle ventricular septal defect was considereddefinitive when colour flow (either turbulent orlaminar) was shown traversing the interventricular

septum. Where no transseptal flow was visualised thepresence of a restrictive defect was inferred if a highvelocity or turbulent jet was identified exiting fromthe right ventricular aspect of the septum into theright ventricular cavity. In restrictive defects thetransseptal and right ventricular flow disturbancewhen visualised was invariably high velocity andturbulent, so a mosaic pattern was seen on thevariance colour flow map (fig 1). In non-restrictivedefects transseptal systolic flow was invariably ofrelatively low velocity and laminar (fig 2) with noturbulence being colour encoded. Transseptal flowwithin the multiple defects was therefore frequentlybest depicted by colour flow imaging in the powermode. ColourM mode studies, in the velocity mode,were also performed in non-restrictive defectsbecause in theory systolic flow acceleration couldoccur either within the defect(s) or at the rightventricle exit point without turbulence being present(fig 3). The inherently higher sampling rate of colourM mode should make this method more sensitive

Fig 1 A modifiedfour chamber view with the transducer angled towards the short axis view and recorded during systole in apatient with-multiple restrictive ventricular septal defects. There are two mosaicjets exitingfrom the right ventricular aspectof the interventricular septum (arrows I and 2). In this view, the mosaic colourflow pattern completely traverses the septumonly through the perimembranous defect (arrow 2). The midtrabecular defect (arrow 1) is depicted by a mosaicjet exitingfrom the right ventricular aspect of the septum. LV, left ventricle; RV, right ventricle; LA, left atrium; RA, right atrium;IVS, interventricular septum.

45

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from

Fig 2 A subcostal view of the right ventricular outflow tract recorded during systole in a patient with multiple non-restrictiveventricular septal defects. Only low velocity laminar flow (blue) with no evidence offlow acceleration traverses theinterventricular septum at the three points at which the ventricular septal defects were present (arrows 1, 2, and 3). There isno aliasing or turbulence within the septum or the right ventricle. See legend tofig 1 for abbreviations.

Fig 3 A colourflow map and colour M mode recorded in the parasternal long axis viewfrom a patient with non-restrictivemultiple muscular ventricular septal defects and an appreciable left to right shunt. The colourflow map on the right isfrozen inlate systole; there is no turbulence within the septum or the right ventricle. In the colour M mode on the left high velocityturbulent diastolicflow entering the left ventricular inflow was caused by high transmitralflow. In early systole there is colourflow traversing the interventricular septum and entering the right ventricle (arrows). The transseptal colourflow shows acolour shiftfrom red to yellow indicatingflow acceleration without turbulence or aliasing. See textforfurther details. aml,anterior mitral valve leaflet. See legend to fig 1 for other abbreviations.

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from

Colourflow imaging in the diagnosis of multiple ventricular septal defects

than colour flow imaging alone and thus could be auseful adjunct in the definition of multiple ven-tricular septal defects by detecting multiple discreteareas of acceleration of transseptal or right ven-tricular flow.

All cardiac catheterisation studies were performedunder general anaesthesia with pressure recordingsbeing taken before angiography. Left ventriculogra-phy was performed in the appropriate angled axialprojections as described by Soto et al."3 Right ven-triculography was also performed in all patientsexcept those with isolated ventricular septal defects.Although both echocardiographic and angio-

graphic studies were sometimes performed by thesame cardiologist in the same patient, all results werelater assessed by one or more experienced indepen-dent observers.Because these patients entered into the study in

various ways it was not always possible to performsimultaneous angiographic and ultrasound studies.But both types of study were always performedwithin 24 hours of each other.The results obtained were initially analysed as a

total patient group and then further analysed afterbeing subdivided into groups with restrictive or non-restrictive defects. Non-restrictive defects weredefined as showing equal peak systolic pressures inboth ventricles measured by cardiac catheterisation.Patients with restrictive multiple ventricular septaldefects had a peak systolic pressure in the leftventricle that exceeded that in the right ventricle,except patients with transposition of the greatarteries in whom the systemic ventricle was on theright.

Results

Cross sectional echocardiography showed at least oneventricular septal defect in all 31 patients and one ormore further defects in 12 (38%) patients; there wasno difference in the detection rate of multiple defectsbetween the groups with restrictive (39%) and non-restrictive defects (38%) respectively (table 2). Thediagnosis ofmultiple ventricular septal defects basedon cross sectional imaging was subsequently con-firmed in all 12 patients by a combination of eithercolour flow imaging, ventriculography, or operationor all three. In this series the addition ofconventionalpulsed and continuous wave Doppler to the crosssectional imaging information did not increase thedetection of multiple ventricular septal defects. Theconventional Doppler methods were useful for dis-tinguishing between patients with restrictive or non-restrictive haemodynamic function. In contrast,colour flow imaging alone, by visualising multipleareas of transseptal flow detected the presence oftwo

Table 2 Diagnostic accuracy of non-invasive and invasivetechniques

Crosssectionalalone CFM LVangioNo (% No(%) No(%)

Restrictive defects(n = 18) 7 (39) 16 (89) 15 (83)

Non-restrictive defects(n = 13) 5 (38) 8 (62) 5 (38)

Total (n = 31) 12 (39) 24 (77) 20 (65)

CFM, colour flow mapping; LV angio, left ventricularcineangiography.

or more ventricular septal defects in 24 (77%) of the31 patients studied. When these were subdivided intogroups with restrictive and non-restrictive defects,colour flow imaging had a much higher diagnosticaccuracy in the restrictive group (89%) than in thenon-restrictive groups (62%). The diagnosis wasconfirmed in all 24 patients by either ventriculo-graphy or operation or both. A comparison of thedifferent colour flow patterns obtained showed thatthere was no difficulty in distinguishing betweenrestrictive (turbulent mosaic pattern) and non-res-trictive (non-turbulent laminar flow pattern) defects.In patients with equal. peak systolic ventricularpressures the detection of multiple ventricular septaldefects depended on the ability to visualise colourencoded transseptal flow. The predominant directionof any shunt did not affect this ability but simplyaltered the colour of the transseptal flow (flowtowards the transducer was encoded as red and flowaway was encoded as blue).Colour M mode studies showed transseptal flow in

the largest defect in patients with non-restrictiveventricular septal defects, but could not confirm thediagnosis ofmultiple defects in any case. Transseptalflow was usually laminar but in one patient systolicflow acceleration was detected. This systolic flowacceleration was indicated by a change in colourintensity while flow itself remained laminar (fig 3).This pattern of flow acceleration correctly predictedthe presence of equal systolic ventricular pressuresand a significant left to right shunt (confirmed atcatheterisation). All other patients with multiplenon-restrictive defects had bidirectional shunting.

Left ventriculography showed multiple ventri-cular septal defects in only 20 (65%) of the 31patients. The diagnostic accuracy of ventriculo-graphy was 83% for restrictive defects and 38% fornon-restrictive defects.

Discussion

The results of this study confirm that colour flow

47

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from

48

imaging is a useful adjunct in the non-invasivediagnosis of multiple ventricular septal defects. Itproved to be better than cross sectional imagingalone, and we found that the addition of the conven-tional Doppler methods did not increase the diagnos-tic yield. Our results for the overall diagnosticaccuracy of cross sectional imaging alone (39%) andcolour flow imaging (77%) for multiple ventricularseptal defects are similar to those previously reportedby Ludomirsky and coworkers who showed anoverall diagnostic sensitivity for cross sectional imag-ing and colour flow imaging of 38% and 72%respectively." When we subdivided our patients intogroups with restrictive and non-restrictive defects,however, colour flow imaging was found to have ahigher sensitivity for the diagnosis of multiple res-trictive defects (89%) than for multiple non-restric-tive defects (62%). This lower sensitivity for non-restrictive defects seems to be a function of theabsence of turbulent transseptal flow, which is aconsequence of the equal peak systolic ventricularpressures. This was the fundamental difference be-tween restrictive and non-restrictive defects asshown by colour flow imaging-turbulent highvelocity flow encoded as a mosaic pattern was easierto visualise than low velocity laminar flow.

Diagnosis in the non-restrictive group was moredifficult when a large ventricular septal defect(usually perimembranous) was found in combinationwith one or more small muscular defects; our studyshowed that under these circumstances colour flowinformation may only be visualised within the largestdefect with no colour flow appearing in the smalldefect(s). This phenomenon was seen with both thevelocity and power display modes. Whether thisrepresented a true lack of flow in the smaller defectsor was an artefact induced by either the analysisalgorithm of the equipment or a failure to aligncorrectly with defect flow is not clear, but if this is atrue finding and flow can occur only over one of themultiple defects present, then it would in part explainthe relatively high incidence of false negative diag-noses when colour flow imaging is used to identifymultiple non-restrictive defects. The inherent highersampling rate of colour M mode might in theorydetect transseptal flow in these small additionaldefects that were missed by colour flow mapping. Inpractical terms, however, this was not the casebecause colour M mode only detected transseptalflow in the largest defect present.The combination of isolated small multiple ven-

tricular septal defects and pulmonary stenosis (orpulmonary band) might make diagnosis difficult,although this particular anatomical arrangement wasnot encountered in this series. Such an arrangementcauses non-restrictive haemodynamic function and

Sutherland, Smyllie, Ogilvie, Keeton

equal peak systolic ventricular pressures, but if thepulmonary obstruction were to be removed then thedefects would be restrictive (because they are small)and might therefore not require concomitant closure.The data described above suggest that such multiplesmall defects will not be reliably visualised by colourflow imaging.There were other difficulties with colour flow

imaging, both in the groups with restrictive and non-restrictive defects when multiple ventricular septaldefects lay close together. The limitation of lateralresolution inherent in the technique do not allow theseparation of turbulent jets that lie within < 0 5 cmofeach other. In contrast, the combination of restric-tive perimembranous and apical trabecular defectswas accurately defined because the turbulent jetswere well separated and there was no difficulty withlateral resolution. Also the detection of multipledefects in the outlet muscular septum (either restric-tive or non-restrictive) will always prove difficultboth because ofthe difficulties in imaging this regionand the subsequent difficulties with lateral resolutionwith colour flow imaging. We believe that these arethe reasons for the diagnostic failures in this study.The overall diagnostic accuracy of left ventriculo-

graphy for multiple ventricular septal defects re-ported in this study (65%) is lower than that reportedby Fellows and coworkers (86%).5 There are prob-ably two reasons for this. Firstly, the proportion ofrestrictive defects in their study group was muchhigher-so the diagnostic accuracy of restrictivedefects in our study (83%) is a more appropriatecomparison and does accord with their findings.Secondly, their study was not primarily designed toassess the diagnostic accuracy of their technique andit is possible that some cases that were missed by leftventriculography were not confirmed by surgicalinspection (that is where a palliative procedure suchas a pulmonary artery band or a systemic-pulmonaryshunt was performed rather than a total correction).This would be particularly applicable to patientswith non-restrictive defects in whom the diagnosisby left ventriculography in this study was especiallypoor (38%).The relative diagnostic accuracy of colour flow

imaging (89%) and left ventriculography (83%) forall types of multiple restrictive ventricular septaldefects in our study was similar. There were no casesin this subgroup that were diagnosed by left ven-triculography but not by colour flow imaging. So leftventriculography was of no additional benefit inpatients in this study with multiple restrictivedefects. In patients with multiple non-restrictivedefects, the results were different; in this subgroupthere were two patients in whom the diagnosis wasmade by left ventriculography but was missed by

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from

Colourflow imaging in the diagnosis of multiple ventricular septal defects 49

colour flow imaging. So although the diagnosticaccuracy for colour flow imaging (65%) was muchhigher than that of left ventriculography (38%), theoverall diagnostic accuracy for multiple non-restric-tive ventricular septal defects was highest when theinformation from the two techniques was usedtogether (77%).None the less, 23% of multiple non-restrictive

defects were misdiagnosed as single isolated ven-tricular septal defects. We feel that this is a fairreflection on the complexity of the diagnosis pre-sented by such defects. It is our opinion that thediagnostic yield will be only slightly increased withmore experience in the use of the diagnostic tech-niques described above and that up to one in five ofmultiple non-restrictive defects and one in 10 ofmultiple restrictive defects may be missed.We conclude that colour flow imaging was the

single most accurate diagnostic technique for theprediction of multiple restrictive ventricular septaldefects. Whereas the diagnosis of multiple non-res-trictive defects was best achieved in this study whencolour flow imaging and left ventriculography wereused together, even the combined approach missed aconsiderable number of patients with multiple ven-tricular septal defects.

References

1 Friedman WF, Mehrizi A, Pusch AL. Multiple mus-cular ventricular septal defects. Circulation 1965;32:35-42.

2 Fox KM, Patel RG, Graham GR, et al. Multiple andsingle ventricular septal defect. A clinical andhaemodynamic comparison. Br Heart J 1978;40:141-6.

3 Breckenbridge IM, Stark J, Waterson DJ, Bonham-

Carter RE. Multiple ventricular septal defects. AnnThorac Surg 1972;13:128-36.

4 Wilkinson JL, Wilcox BN, Anderson RH. The anatomyof double outlet right ventricle. In: Anderson RH,Macartney FJ, Shinebourne EA, Tynan M, eds.Paediatric cardiology Vol 5. Edinburgh: ChurchillLivingstone, 1981:397-407.

5 Fellows KE, Westerman MD, Keane JF. Angiographyof multiple ventricular septal defects in infancy.Circulation 1982;66:1094-9.

6 Soto B, Bargeron LM, Pacifico AD, Vanini V, KirklinJW. Angiography of atrioventricular canal defects.Am J Cardiol 1981;48:492-9.

7 Kirklin JK, Castaneda AR, Keane JF, Fellows KE,Norwood WI. Surgical management of multipleventricular septal defects. J Thorac Cardiovasc Surg1980;80:485-93.

8 Taylor JFN, Chrispin AR. Interventricular septaldefect shown by left ventricular cine-angiocardio-graphy. Br Heart J 1971;33:285-9.

9 Sutherland GR, Godman MJ, Smallhorn JF, GuiterrasP, Anderson RH, Hunter S. Ventricular septaldefects. Two dimensional echocardiographic andmorphological correlations. Br Heart J 1982;47:316-28.

10 Magherini A, Simonetti L, Tomassini CR, Moggi C,Ragazzini F, Bartolozzi G. Cross-sectional echocar-diography with pulsed and continuous wave Dopplerin the management of ventricular septal defects. Int JCardiol 1987;15:317-28.

11 Ludomirsky A, Huhta JC, Vick GW, Murphy DJ,Danford DA, Morrow WR. Color Doppler detectionof multiple ventricular septal defects. Circulation1986;74:1317-22.

12 Ortiz E, Robinson PJ, Deanfield JE, Franklin R,Macartney FJ, Wyse RKH. Localisation of ven-tricular septal defects by simultaneous display ofsuperimposed colour Doppler and cross sectionalechocardiographic images. Br Heart J 1985;54:53-60.

13 Soto B, Coghlan CH, Bargeron LM Jr. Angiography inventricular septal defects. In: Anderson RH,Shinebourne EA, eds. Paediatric cardiology.Edinburgh: Churchill Livingstone, 1978:125-35.

on 24 April 2019 by guest. P

rotected by copyright.http://heart.bm

j.com/

Br H

eart J: first published as 10.1136/hrt.62.1.43 on 1 July 1989. Dow

nloaded from