-
Archives ofDisease in Childhood 1996;75:416-422
Early identification of anthracyclinecardiomyopathy:
possibilities and implications
Frances A Bu'Lock, Martin G Mott, Anthony Oakhill, Robin P
Martin
AbstractThe number of survivors of childhoodcancer affected by
anthracycline cardio-myopathy is steadily increasing,
despiteefforts to limit cardiotoxicity by doserestriction. Cardiac
function was evalu-ated prospectively in 125 children
duringtreatment to attempt to identify individualsusceptibilities
to cardiotoxicity andhence any potential for treatment
modifi-cation.
Left ventricular shortening fraction wasused as an index of
cardiotoxicity. Short-ening fraction declined as
cumulativeanthracycline dose increased, at an aver-age rate of 1%
per 100 mg/m'. Six patients(5%) developed heart failure. Twenty
fourpatients (19%) had abnormal shorteningfraction (200 mg/m'.
Patients at high risk ofrisk ofimportantanthracycline
cardiotoxicity may be iden-tifiable early in treatment by regular
andcareful monitoring of shortening fraction.However, frequent
assessment is requiredand this has significant resource
implica-tions.(Arch Dis Child 1996;75:416-422)
Keywords: cardiomyopathy, anthracycline,
monitoring,echocardiography.
Bristol Royal Hospitalfor Sick Children,Department ofPaediatric
CardiologyFA Bu'LockRP Martin
Department ofPaediatric OncologyMG MottA Oakhill
Correspondence to:Dr F A Bu'Lock,Department of
PaediatricCardiology, Alder HeyChildren's Hospital, EatonRoad,
Liverpool Ll 2 2AP.
Accepted 6 June 1996
The number of children and young adultsdeveloping cardiomyopathy
after chemo-therapy for childhood malignancy is steadilyincreasing.
Of the 1300 new cases of childhoodmalignancy in the UK each year,
more thanhalf will receive cardiotoxic agents. For the last10-15
years, around 70% of patients have beenachieving long term
survival,' thus >5000 sur-vivors are already at risk in the UK
alone.The principal cardiotoxic agents used in
chemotherapy are the anthracyclines doxoru-bicin, daunorubicin,
and epirubicin, whichcause a dose related, cumulative loss of
cardiacmyocytes,2' probably mediated by toxic freeradicals
generated by an intracellular iron-anthracycline complex.56 Cardiac
irradiation,incurred during thoracic, craniospinal, or total
body radiotherapy will also cause significantcardiac damage,
which is exacerbated byconcomitant anthracycline treatment."
Highdose cyclophosphamide91' and mito-zantrone" 12 also cause
cardiac injury.A number of ways to reduce cardiotoxicity
have been suggested but the use of less toxicanthracyclines,"'4
prolonged infusion sched-ules,5 16 or adjuvant cardioprotective
agents,7 18although increasingly advocated, are not yet ofproved
utility. Meanwhile, further attempts toimprove rates of cure using
'megatherapy' maysucceed only at the expense of an increase inthe
incidence of anthracycline cardiomyopathy.
Clinical evidence of anthracyclinecardiotoxicity-congestive
cardiac failure-isonly manifest very late. By this stage,
histologi-cal myocardial damage is severe, all physiologi-cal
compensatory mechanisms have failed, andthe prognosis is very poor
indeed.'6.19.20 Histo-logical monitoring of subclinical cardiac
dam-age is possible using electron microscopicassessment of serial
endomyocardial biopsyspecimens,2 22 but this is both invasive
andhighly resource intensive. Thus it is rarelyapplicable in
clinical practice and measures ofleft ventricular function are
generally used asless invasive surrogate markers of
subclinicalmyocardial damage. Overall left ventricularejection
function can be assessed using radio-nuclide angiography2 24 or
echocardiogra-phy2526 and some groups have advocatedindices related
to 'myocyte contractility', whichmay also be obtained
echocardiographi-cally.2728 Measurement of transverse left
ven-tricular systolic shortening fraction from the Mmode
echocardiogram is probably the mostwidely used method. Although,
like all non-invasive techniques, the use of shortening frac-tion
does have some limitations, it has beenwell validated as a
surrogate index of anthracy-cline cardiotoxicity by many groups.The
majority of studies reporting the effects
of anthracyclines on left ventricular functionhave been cross
sectional and have generallyconcentrated on status after treatment.
How-ever, most report a relationship betweendecline in left
ventricular function and increas-ing anthracycline dose, although
with themarked individual variability in dose sensitivityalso noted
in both clinical'29'0 and histopatho-logical" 2 studies. It is this
individual variabil-ity that imposes limits on the efficacy
ofuniform dose restriction in preventing anthra-cycline
cardiomyopathy. However, if also mani-fest during treatment, it
offers the possibilityfor early identification of susceptibility
tocardiotoxicity to allow modification of subse-
416
on July 8, 2021 by guest. Protected by copyright.
http://adc.bmj.com
/A
rch Dis C
hild: first published as 10.1136/adc.75.5.416 on 1 Novem
ber 1996. Dow
nloaded from
http://adc.bmj.com/
-
Early identification of anthracycline cardiomyopathy
quent treatment in the particularly 'atrisk'patient.The results
of a detailed prospective study of
left ventricular function in children receivinganthracycline
treatment were therefore exam-ined to determine whether the
variation in sus-ceptibility to anthracycline cardiotoxicity
couldbe detected using left ventricular shorteningfraction. If so,
it was hoped to determine crite-ria that might be used for
individual tailoringof antitumour treatment on the basis of
bothoncological and cardiac status.
Subjects and methodsBetween August 1989 and July 1992,
allchildren treated in our unit with chemotherapyregimens including
anthracyclines participatedin a prospective cardiotoxicity study.
Allpatients underwent detailed cardiovascular andechocardiographic
examinations before first'on study' anthracycline treatment and
shortlybefore each subsequent anthracycline treat-ment, as well as
one month after the last treat-ment.
Height and weight were measured on eachoccasion and body surface
area (BSA, in m')calculated from standard nomograms. Bloodpressure
was recorded in mm Hg and anyabnormal cardiovascular symptoms or
signswere noted. Blood haemoglobin concentrationwas >80 g/l on
all occasions. All echocardio-graphic examinations were performed
andanalysed by a single experienced operator(FAB).
Left ventricular systolic function was as-sessed by conventional
steered M modeechocardiography from the parasternal longaxis view
using a Hewlett Packard 77020ACultrasound scanner with simultaneous
electro-cardiographic recording. Left ventricular di-astolic
diameter (LVDD) and posterior walldiastolic thickness were measured
at the pointof maximum diastolic posterior deflection ofthe
posterior wall. Left ventricular systolicdiameter (LVDS) and
posterior wall systolicthickness were measured at the point of
maxi-mum anterior deflection of the posterior wall.The mean of
measurements from four cardiaccycles were used and left ventricular
shorten-ing fraction (%) was calculated from the meanleft
ventricular dimensions as:
Shortening fraction = (LVDD - LVDS)x 100.LVDD
STUDY GROUPDuring the study period, 138 children under-went a
total of 651 serial cardiovascular andechocardiographic
examinations during an-thracycline chemotherapy for a variety
ofhaematological and solid malignancies. An-thracyclines were
administered by slow bolusinjection in increments of 35-100 mg/m'
andfour patients also received cardiac irradiation.Where patients
received additional cardiotoxicbut non-anthracycline treatment,
subsequentexamination details have not been included.
Studies were stratified by cumulative anthra-cycline dose into
nine groups: pretreatment,1-100 mg/M2, 101-200 mg/m', etc up to
700
mg/m', and >700 mg/m'. Equipotency for car-diotoxicity was
assumed for all three types ofanthracycline (see below). Where more
thanone study per patient fell within a single dosegroup, only the
study performed at the highestdose was used for analysis.Data from
13 patients were completely
excluded by reason of death during inductiontreatment (n = 3),
experimental cardioprotec-tion (n = 4), structural cardiac
abnormalities(n = 2), technical difficulties (n = 1), or
serialstudies only within the same 100 mg/m' doserange (n = 3).
After data stratification, a total of 451 serialechocardiograms
performed during treatmentof 125 children for malignant disease
wereused for analysis. There were 66 boys and 59girls, with BSAs
between 0.34 and 2.08(median 0.8) m'. They were aged between 0.5and
20 (median 6.3) years at first study andreceived cumulative
anthracycline doses be-tween 45 and 1150 (mean 320, median
270)mg/m'. Ninety two patients were studiedbefore receiving any
anthracycline, 33 hadreceived prior anthracyclines and their
dataseries started at doses between 70 and 600(median 150)
mg/M2.
CONTROL DATAStudy data were compared with those obtainedfrom 137
healthy children and young adultsexamined by identical techniques,
expressed asgroup median values for 0.2 m' BSA inter-vals.3334
Matched median normal control val-ues of all study parameters were
determinedfor each patient on the basis of their BSA atfirst
examination and were used for pairedcomparisons as described
previously else-where."4
In our control studies fractional shorteningwas found to be
independent of body size orage. The normal values were 38% for the
50thcentile, 35% for the 25th, and 34% for the1 0th. No normal
subject had a shortening frac-tion 35% was consideredunequivocally
normal, 30-34% 'borderline',and
-
Bu'Lock, Mott, OakhiU, Martin
Table 1 Changes in median values ofparameters of left
ventricular systolic function with increasing anthracycline
dose
Dose groupPatient (P) orcontrol (C) data 0 1 2 3 4 5 6 7 8
END
Dose (mg/m2) P 0 90 180 285 385 483 573 686 857 270Age (years) P
6.31 7.38 7.27 7.95 8.08 9.29 7.87 8.09 8.48 7.58
C 5.5 5.5 5.5 5.5 5.5 8.01 8.01 5.5 5.5 5.5BSA (m') P 0.81 0.85
0.85 0.94 0.90 0.99 0.98 0.94 1.0 0.88
C 0.78 0.78 0.78 0.78 0.78 1.02 1.02 0.78 0.78 0.78BPS (mmHg) P
110 110 100 100 105 100 100 85 90 100
C 95 95 95 95 95 95 95 95 95 95BPD (mm Hg) P 70 67 60 65 60 60
60 60 60 60
C 63 63 63 63 63 63 64 63 63 63LVDD (cm) P 3.64 3.66 3.73 3.80
3.82 3.88 3.78 3.81 4.26 3.80
C 3.63 3.63 3.63 3.63 3.63 3.91 3.91 3.63 3.63 3.63LVDS (cm) P
2.34 2.37 2.40 2.46 2.52 2.58 2.58 2.66 3.19 2.46
C 2.29 2.29 2.29 2.29 2.29 2.47 2.47 2.29 2.29 2.29SF(%) P 38 36
36 34 33 33 32 29 27 34
C 38 38 38 38 38 38 38 38 38 38PWD (cm) P 0.06 0.63 0.60 0.64
0.65 0.63 0.66 0.60 0.65 0.62
C 0.56 0.56 0.56 0.56 0.56 0.60 0.60 0.56 0.56 0.56PWS (cm) P
0.97 1.0 0.95 1.0 0.97 1.05 .98 0.89 0.92 0.96
C 0.92 0.92 0.92 0.92 0.92 1.03 1.03 0.92 0.92 0.92% withSF
-
Early identification of anthracycline cardiomyopathy
0.0008 0.0003
T ENDSF.-AI>30%
301-4l0 401-1u0
Figure 2 Shows median and interquartile range (error bars)
values for left ventricularshorteningfraction with increasing
anthracycline dose, subdivided on the basis ofend oftreatment
shortening fraction (ENDSF 30%, n = 101, and ENDSF
-
Bu'Lock, Mott, Oakhill, Martin
first studied before receiving any anthracycline,11 had end of
treatment shortening fraction
-
Early identification of anthracycline cardiomyopathy 421
cumulative anthracycline dose and left ven-tricular shortening
fraction. The marked indi-vidual variability in the rate of fall of
fractionalshortening with increasing anthracycline doseis
consistent with other studies.'29-32 It is there-fore highly
noteworthy that each individual'sbehaviour appears to be relatively
consistentthroughout treatment and has a clear relation-ship to the
functional status at the end ofanthracycline treatment. Thus, the
use of acombination of measurement of the rate of fallof shortening
fraction and of individual abso-lute values of fractional
shortening at lowanthracycline doses, should allow identifica-tion
of patients particularly susceptible toanthracycline cardiotoxicity
early on in treat-ment. However, these results are drawn from
aretrospective analysis of prospectively acquireddata and would
benefit from further prospec-tive validation.The most disturbing
aspects of anthracy-
cline induced cardiac damage are the long termimpairment of
cardiac function and the in-creasing incidence of late
decompensationwhich have only recently become appar-ent.283437 It
appears that after an initial 'honey-moon period',"839 in which
surviving myocytesare able to compensate for the acute myocyteloss
by hypertrophic changes (fractional short-ening may even return to
near normal values),there is a continuing myocardial
attenuation,probably due both to failure of further myocytegrowth
and gradual loss of 'overworked'residual myocytes. Our own studies
and thoseof other groups have now demonstrated thatmyocardial
thickness and function deterioratewith time after anthracycline
treatment,283437with more than 40% of patients showingsignificant
abnormalities of cardiac function
115 years after anthracycline treatment.40Since Steinherz et al
have also demonstrated aclear relationship between end of
treatmentcardiac status/shortening fraction and likeli-hood of late
decompensation"37 the ability topredict likely functional outcome
early in treat-ment becomes even more crucial.There has been
considerable discussion
about both the value and timing of monitoringfor anthracycline
cardiotoxicity.2'26130 41 Theguidelines suggested by the Children's
CancerStudy Group were based on a large experienceat the Memorial
Sloan Kettering CancerCenter and on previously reported
studies.42These guidelines have been criticised byLipshultz et
al,4' as being both unproved and ashaving potential negative
consequences forantitumour treatment. The present 'on treat-ment'
study is the first to identify a number ofpotential pointers to
each individual's suscepti-bility to anthracycline cardiotoxicity,
which canbe applied throughout treatment. It is not sug-gested that
these criteria be used to define rigidlimits for cessation of
anthracycline treatment,but rather to provide the basis for
informedconsideration of the risks and benefits ofcontinuation of
anthracycline treatment in anygiven oncological situation. Overt
cardiacfailure is an absolute contraindication tocontinuation of
anthracycline treatment; lesserdegrees of myocardial dysfunction or
increased
sensitivity to cardiotoxicity may warrant eitherdose reduction
or early cessation of treatment.Where the risk that treatment
modificationmay significantly jeopardise antitumour suc-cess is
substantial, an informed decision tocontinue anthracyclines may be
justified. Inless susceptible patients, the beneficial use ofhigher
than usual doses of anthracycline maybe facilitated and the
standard recommendedmaximum doses perhaps need no longer apply.As
cardioprotective agents become avail-able,"'5 identification of
those patients mostsusceptible to cardiotoxicity might provide
arational basis for their selective use.
ConclusionsRegular monitoring of left ventricular shorten-ing
fraction early during anthracycline treat-ment can identify those
patients at higher riskof subsequent cardiotoxicity. Patients
withfractional shortening 2-3 absolute % per 100 mg/m'would appear
to be at increased risk of signifi-cant cardiotoxicity.
Modification of treatmentshould be considered for these patients in
thelight of each individual's tumour status.The echocardiographic
determination of
individual susceptibility to anthracycline car-diotoxicity
requires pretreatment and frequentlow dose studies, and a
meticulous anduniform echocardiographic technique. Suchmonitoring
has significant resource implica-tions. Further prospective studies
of theclinical value and cost effectiveness of on treat-ment
monitoring are now indicated, using thecriteria determined in this
study.
FAB was supported by the Cancer Research Campaign, MGMis
professor ofpaediatric oncology and is supported by the Can-cer and
Leukaemia in Childhood Trust.
1 Mott MG. Neoplasia in childhood: 25 years of progress.Ann
Oncol 1995 (in press).
2 Young RC, Ozols RF, Myers CE. The anthracyclineanti-neoplastic
drugs. NEnglJMed 1981;305:139-53.
3 Praga C, Beretta G, Vigo PL, et al. Adriamycincardiotoxicity:
a survey of 1273 patients. Cancer Treat Rep1979;63:827-34.
4 Myers CE, McGuire WP, Liss RH , et al. Adriamycin; therole of
lipid peroxidation in cardiac toxicity and tumourresponse. Science
1977;197:165-7.
5 Doroshow JH, Reeves J. Anthracycline enhanced oxygenradical
formation in the heart. Proceedings of the AmericanAssociation For
Cancer Research 1980;21:266(abstr).
6 Lefrak E, Pitha J, Rosenheim S, Gottlieb JA. A
clinicopatho-logic analysis ofAdriamycin cardiotoxicity. Cancer
1973;32:302-14.
7 Billingham ME, Bristow MR, Glatstein E, et al.
Adriamycincardiotoxicity, endomyocardial biopsy evidence of
en-hancement by irradiation. Am J Surg Pathol 1977;1:17-23.
8 Fajardo LF, Eltringham JR, Stewart JR. Combined
cardio-toxicity of adriamycin and X-radiation. Lab Invest
1976;34:86-96.
9 Mills BA, Roberts RW.Cyclophosphamide inducedcardiomyopathy: a
report of two cases and a review of theEnglish literature. Cancer
1979;43:2223-6.
10 Braverman AC, Antin JH, Plallert MT, Cook EF, Lee
RT.Cyclophosphamide cardiotoxicity in bone-marrowtransplantation: a
prospective evaluation of new dosingregimens. J Clin Oncol
1991;9:1215-23.
11 Unverferth DV, Unverferth BJ, Balcerzak SP, Bashore
TA,Neidhart JA. Cardiac evaluation of mitoxantrone. CancerTreat Rep
1983;67:343-50.
12 Pratt CB, Crom DB, Wallenberg J, et al. Fatal congestiveheart
failure following mitoxantrone therapy in twochildren previously
treated with doxorubicin and cisplatin.Cancer Treat Rep
1983-67:85-8.
13 Nielsen D, Jensen JB, Dombernowsky P, et al.
Epirubicincardiotoxicity: a study of 135 patients with advanced
brastcancer.JClin Oncol 1990;8:1806-10.
on July 8, 2021 by guest. Protected by copyright.
http://adc.bmj.com
/A
rch Dis C
hild: first published as 10.1136/adc.75.5.416 on 1 Novem
ber 1996. Dow
nloaded from
http://adc.bmj.com/
-
422 Bu'Lock, Mott, Oakhill, Martin
14 Bonnadonna G, Gianni L, Santoro A, et al. Drugs ten
yearslater: epirubicin. Ann Oncol 1993;4:359-69.
15 Legha SS, Benjamin RS, Mackay B, et al. Reduction
ofdoxorubicin cardiotoxicity by prolonged continuousintravenous
infusion. Ann Intern Med 1982;96: 133-9.
16 Casper ES, Gaynor JJ, Hajdu SI, et al. A
prospectiverandomised trial of adjuvant chemotherapy with bolus
ver-sus continuous infusion of doxorubicin in patients withhigh
grade extremity soft tissue sarcoma: an analysis ofprognostic
factors. Cancer 1991;68:1221-9.
17 Speyer JL, Green MD, Zeleniuch-Jaquotte A, et al. ICRFpermits
longer treatment with doxorubicin in women withbreast cancer. Jf
Clin Oncol 1992;10: 117-27.
18 Bu'Lock FA, Gabriel HM, Oakhill A, Mott MG, Martin
RP.Cardioprotection by ICRF187 against high dose anthracy-cline
toxicity in children with malignant disease. Br Heart31993;70:
185-8.
19 Von Hoff DD, Rozencweig M, Layard MW, Slavik M, Mug-gia FM.
Daunomycin-induced cardiomyopathy in childrenand adults. Am J'Med
1977;62:200-8.
20 Dearth J, Osborn R, Wilson E, et al.
Anthracycline-inducedcardiomyopathy in children. A report of six
cases. MedPediatr Oncol 1984;12:54-8.
21 Billingham ME, Mason JW, Bristow MR, Daniels JR.Anthracycline
cardiomyopathy monitored by morphologicchanges. Cancer Treat Rep
1978;62:865-72.
22 Torti FM, Bristow MM, Lum BL, et al. Cardiotoxicity
ofepirubicin and doxorubicin: assessment by endomyocardialbiopsy.
Cancer Res 1986;46:3722-7.
23 Palmeri ST, Bonow RO, Myers CE, et al. Prospective
evalu-ation of doxorubicin cardiotoxicity by rest and
exerciseradionuclide angiography. Am J Cardiol 1986;58:607-13.
24 Schwartz RG, McKenzie WB, Alexander J, et al. Congestiveheart
failure and left ventricular dyshortening fractionunc-tion
complicating doxorubicin therapy: seven yearexperience using serial
radionuclide angiography. Am JMed 1987;82:1109-18.
25 Mott MG, Jordan SC.Cardiotoxicity of adriamycin assessedby
echocardiography. Proceedings of the 10th Meeting ofSocieti
Internationale Oncologie Pediatrique. Belgium 1978:188 (abstr).
26 Bloom KR, Bini RM, Williams CM, Sonley MJ, GribbinMA.
Echocardiography in adriamycin cardiotoxicity. Can-cer
1978;41:1265-9.
27 Colan SD, Borow KM, Neumann A. Left ventricularend-systolic
wall stress-velocity of fiber shortening relation:a load
independent index of myocardial contractility. J AmCoil Cardiol
1984;4:715-24.
28 Lipshultz SE, Colan SD, Gelber RD, et al. Late cardiaceffects
of Adriamycin therapy for childhood acute lym-phoblastic leukaemia.
N Eng3 Med 199 1;324:808-15.
29 Bristow MR, Thompson PD, Martin RP, et al. Early
anthra-cycline cardiotoxicity. Am J Med 1978;165:823-32.-
30 Minow RA, Benjamin RS, Lee ET, Gottleib JA.
Adriamycincardiomyopathy: risk factors. Cancer
1977;39:1397-402.
31 Isner JM, Ferrans VJ, Cohen, SR, et al. Clinical
andmorphological findings after anthracycline chemotherapy.Am
ICardiol 1983;51:1167-74.
32 Bristow MR, Lopez MB, Mason JW, Billingham ME, Win-chester
MA. Efficacy and cost of cardiac monitoring inpatients receiving
doxorubicin. Cancer 1982;50:32-41.
33 Bu'Lock FA, Mott MG, Oakhill A, Martin RP. Leftventricular
diastolic function in children measured byDoppler echocardiography:
normal values and relationwith growth. BrHeartJ 1995;73:334-9.
34 Bu'Lock FA, Mott MG, Oakhill A, Martin RP. Leftventricular
diastolic function after anthracycline chemo-therapy in childhood:
relationships with systolic function,symptoms and pathophysiology.
Br Heart J 1995;73:340-50.
35 Von Hoff DD, Layard MW, Basa P, et al. Risk factors
fordoxorubicin-induced congestive heart failure. Ann InternMed
1979;91:701-17.
36 Lipshultz SE, Lipsitz SR, Mone SM, et al. Female sex
andhigher drug dose as risk factors for late cardiotoxic effectsof
doxorubicin therapy for childhood cancer.
NEnglJMed1995;332:1738-43.
37 Steinherz L, Steinherz P, Tan CTC, Heller G, Murphy
ML.Cardiac toxicity 4-20 years after completing
anthracyclinetherapy. JAMA 1991;266:1672-7.
38 Lewis AB, Crouse VL, Evans W, Takahashi M, Siegel SE.Recovery
of left ventricular function after discontinuationof anthracycline
chemotherapy in children. Pediatrics 1981;68:67-72.
39 Saini J, Rich MW, Lyss AP. Reversibility of severe
LVdysfunction due to doxorubicin cardiomyopathy; report ofthree
cases. Ann Intern Med 1987;106:814-6.
40 Steinherz LJ, Steinherz PG, Sklar C, Wollner N, Tan C.Cardiac
status of 42 patients > 15 years post anthracyclinetherapy. Med
Pediatr Oncol 1994;23:176 (abstr).
41 Mason JW, Bristow MR, Billingham ME, Daniels JR. Inva-sive
and non-invasive methods of assessing Adriamycincardiotoxic effect
in man: superiority of histopathologicassessment using
endomyocardial biopsy. Cancer Treat Rep1978;62:857-64.
42 Steinherz U, Graham T, Hurwitz R, et al. Guidelines
forcardiac monitoring of children during and after anthracy-cline
therapy: report of cardiology committee of Children'sCancer Study
Group. Pediatrics 1992;89:942-9.
43 Lipshultz SE, Sanders SP, Goorin AM, Krischer JP, SallanSE,
Colan SD. Monitoring for anthracycline cardiotoxicity.Pediatrics
1994;93:433-7.
on July 8, 2021 by guest. Protected by copyright.
http://adc.bmj.com
/A
rch Dis C
hild: first published as 10.1136/adc.75.5.416 on 1 Novem
ber 1996. Dow
nloaded from
http://adc.bmj.com/