Complication Probability as Assessed from Dose-Volume Histograms

Author(s): John T. Lyman

Source: Radiation Research Supplement, Vol. 8, Heavy Charged Particles in Research and
Medicine. Proceedings of a Symposium Held at the Lawrence Berkeley Laboratory, University
of California, Berkeley, California, May 1-3, 1985 (Nov., 1985), pp. S13-S19
Published by: Radiation Research Society
Stable URL: http://www.jstor.org/stable/3583506 .
Accessed: 20/02/2015 13:02
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Radiation Research Society is collaborating with JSTOR to digitize, preserve and extend access to Radiation
Research Supplement.

http://www.jstor.org

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

RADIATION
RESEARCH
104, S-13-S-19 (1985)

ComplicationProbabilityas Assessed from

Dose-Volume Histograms
JOHN T. LYMAN
LawrenceBerkeleyLaboratory,Berkeley,California94720
as AssessedfromDose-VolumeHistograms.Radiat.
LYMAN,J. T. ComplicationProbability
Res., Suppl.8, 104, S-13-S-19(1985).
Optimizationof a treatmentplan for radiationtherapywill producea plan with the highest
probabilityfortumorcontrolwithoutexceedingan acceptablecomplicationrate.To achievethis
goal it is necessaryto havea meansto estimateprobabilitiesof localcontroland normaltissue
complication.In general,goodtreatmentplansdelivera highuniformdose to the targetvolume
andlowerdosesto thesurrounding
normaltissues.Thetolerancedosevaluesavailableforvarious
normaltissuesare usuallyassumedto applyto partialor full volumesof the tissuewhichhave
beenuniformlyirradiated.
Thesevaluesarethe bestguidelinesforestimatingcomplicationprobabilitiesin tissuesthat receivea uniformdose to a fractionof the tissue and no dose to the
remainder.
Dose-volumehistograms
areone meansof evaluatingthe uniformityof the irradiation
on the tissues.Frequentlythe normaltissuesarenot uniformlyirradiatedas is demonstrated
by
dose-volumehistogramsfor differenttreatmentplans.A recursivealgorithmwhichuses these
dose-volumehistogramsto
tolerancedosedatahasbeenwrittenand can be appliedto arbitrary
estimatethe complicationprobability.
INTRODUCTION

Treatment planning optimization involves selection of an acceptable plan that controls the local and regional spread of cancer without causing normal tissue complications (1). To fully optimize a treatment plan (maximize the tumor dose without
exceeding an acceptable complication probability), a method is needed to determine
the complication probabilities for the different normal tissues which will be affected
by the treatment.
Sophisticated treatment planning, particularly for particle beam therapy, involves
three-dimensional calculations or a series of two-dimensional calculations using adjacent slices of CT data (2). To evaluate a given plan, the radiotherapistmust integrate
the information presented by series of isodose distributions superimposed on the CT
images. All the slices, which demonstrate the target volume and the critical tissues,
must be examined to ascertain that the target is adequately treated and the normal
tissues are adequately spared. Heavy charged-particlebeams offer a means for sparing
normal tissues from irradiation that is not available with conventional radiations.
Decisions to maximally spare normal structures must be tempered by the possibility
of not adequately treating normal tissues bearing occult disease.
If dose-volume histograms are calculated from the data presented in the isodose
distributions, the degree of uniformity or nonuniformity of the irradiation may be
S-13

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

0033-7587/85$3.00

S-14

JOHN T. LYMAN

quickly assessed because a single histogram can present dose uniformity information
which is contained in a number of sets of isodose distributions (3).1'2 The histograms
give another way to look at the plan and may give some additional information that
is not readily apparent when using the isodose distributions. A high-dose region in
the histogram may represent a high-dose region that is related to a single CT slice or
may represent a contiguous region related to adjacent CT slices or even a number of
isolated high-dose regions related to the same or different slices. The isodose distributions are needed to determine the location of the high-dose regions. Therefore the
dose-volume histogramsdo not substitutefor the isodose distributionsin the evaluation
of a treatment plan, but augment them.
The purpose of this study is to develop a method to estimate the complication
probability of a normal tissue structure from a dose-volume histogram and data from
a population of radiation therapy patients. The estimates of the complication probability can then be used to optimize a treatment plan. The identification of the optimal
plan is also subject to clinical judgment of the acceptability of the plan and the risk/
benefit for each individual patient. This approach is being explored in the optimization
of heavy-charged-particle treatment plans. An evaluation of several representative
cases has led to some generalized conclusions which may provide guidance to the
radiotherapist in assessing appropriate treatment plans. The techniques developed in
this work are equally applicable to optimization of treatment plans for other radiations.
METHODS
Datafromthe literaturewhichmightaid in thisrankingarevaluesof the tolerancedoseof varioustissues
(4-6). Tolerancedosesare usuallygivenas a functionof beamarea,or lengthof the irradiatedportionof
structure,or fractionof the organtreated.Thesedata,whicharepartof the folkloreof radiotherapy,
arefor
treatmentwith low-LETradiationsand a conventionalfractionschedule.Thesevalueswerederivedfrom
observationof patientswho weretreatedbeforethe daysof sophisticatedtreatmentplanning.Thereforeit
mustbe assumedthattherecouldbe significanterrorsin the estimationof theportionof theorganortissue
which receivedthe irradiationand in the uniformityof the irradiation.These data, which are the best
available,are not as completeor as firmas desirable.
The bestexampleof the tolerancedose (in Gy) is forthe heart(TableI). The dataforthe partial-volume
irradiations
arefromRubinet al. (5), whilethe remainingpointrepresents
thecollectivejudgmentof several
experiencedtherapistswho are participatingin an NCI-sponsored
projectto evaluateparticletreatment
plans.The TD50and TD5are the dosesthat wouldresultin 50 and 5%complicationprobabilitiesafter5
years,respectively,and representtwo points on what is assumedto be a sigmoid-shaped
dose-response
curve.These data imply that the complicationprobabilityis a functionof both the percentagevolume
irradiatedand the absorbeddose receivedby the volume.Thereare manyotherfactorsthat are involved,
most noticeablythe fractionationscheme;however,forthe momentthisand all theseotherfactorswill be
heldconstant.
Datafor otherpartialvolumesand tolerancelevelscan be obtainedby interpolationor extrapolation.
A
fullsetof datacanbe represented
surfacethatrepresents
theprobability
of complication
by a three-dimensional
as a functionof both the volumeand the dose (Fig. 1).This surfacewasdeterminedby assumingthat the
volumedependencecouldbe represented
by a power-lawrelationship(7, 8).
'S. R. Zink,J. R. Castro,G. T. Y. Chen,J. M. Collier,J. T. Lyman,and W. M. Saunders,Treatment
withphotonsforcarcinomaof the esophagus.Manuscript
planningstudycomparesheavyion radiotherapy
submittedto Int.J. Radiat.Oncol.Biol. Phys.
2 M. M. Austin-Seymour,
G. T. Y. Chen,J. R. Castro,W. M. Saunders,S. Pitluck,K. H. Woodruff,and
M. Kessler,Dose volumehistogramanalysisof liverradiationtolerance.Manuscriptsubmittedto Int. J.
Radiat.Oncol.Biol. Phys.

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

S-15

COMPLICATION
PROBABILITY
TABLEI
ToleranceDosesfor Heart
% Volume

TDso

TD5

25
60

80
55

70
45

35

100

TD(V) TD(I)/j",

(1)

whereTD(V) is the tolerancedose for a givenpartialvolume(V), TD( ) is the tolerancedose for the full
is represented
Thedosedependence
volume,andn is a fittedparameter.
bytheintegralof a normaldistribution
of a sigmoidcurve).
(one of severalpossiblerepresentations
PC
=

e-'2/2dt,
-

(2)

where t = (D - TD5o(V))/a( V)). This curve is completely defined by the mean (TD50(V)) and the standard

deviation(a(V))whichforthecurrentworkhasbeenapproximated
by m X TD5( V).Thethree-dimensional
surfacecan then be completelydefinedby the threeparameters,
TD5o(), n, and m.
tolerancedoses for the heart(in Gy) calculatedwith Eqs.(1)
Table II is an exampleof partial-volume
and (2) and parametersderivedfromthe data in TableI. Calculatedvaluesfor pointsthat correspondto
valuesin TableI are underlined.Valuesof TD5o(1),n, and m for a selectedlist of organsand tissuesare
given in TableIII.Thesevaluesare preliminaryestimatesbasedon publishedand unpublisheddata and
willemergefrom
opinions;as such,theyshouldbe usedwithcaution.It is anticipatedthatbetterparameters
treatmentplanningsystemsanddose-volumehistograms.
studiesusingthree-dimensional
Cumulativedose-volumehistogramsfora kidneyfor severalpossibletreatmentplansareshownin Fig.
2. Thesehistogramsshowthe fractionof the organwhichreceivesa dose equalto or greaterthana given
value.The histogramswerederivedfromtreatmentplansusinghelium,carbon,or neon beamswith field
all beamsproducedessentiallythe
of two, three,or fourfields.Forthe two-fieldarrangements
arrangements
samehistogram,the one withthe largestdose.Thefour-fieldarrangement
givesthe lowestdoseswithhelium
withthethree-field
occurs
This
same
neon.
which
is
lower
than
lower
than
carbon
arrangements.
ranking
being
andeventherelativerankingof thesehistograms
It is notdifficultto determinethe mostdesirabledistribution
by visualinspection.The averagedoseor the integraldosecan alsobe usedto aid in the rankings.However,
betweenthehistograms.
of thedifferences
thesemethodsdo notprovidean estimateof theclinicalsignificance
Providedthereare sufficientdatato obtainan estimateof the three-dimensional
complicationsurface,an
estimateof the probabilityfor complicationcan also be obtained,sincethesehistogramsrepresenta fairly

PC0.5-

0
1.0
Partial
0.5
Volume

\D 20

v
4U
ose (Gray)
Dose (Gray)

of the probabilityof complicationfor the heartas a

surfacerepresentation
FIG.1. A three-dimensional
function of the dose and the partial volume which is uniformily irradiated.

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

S-16

JOHN T. LYMAN
TABLEII
DerivedToleranceDosesfor Heart
V

TDso(V)

TD5(V)

0.200
0.250
0.300
0.350
0.400
0.450
0.500
0.550
0.600
0.650
0.700
0.750
0.800
0.850
0.900
0.950
1.000

93.7
83.8
76.5
70.8
66.2
62.5
59.5
56.5
54.1
52.0
50.1
48.4
46.9
45.5
44.2
43.0
41.9

78.3
70.0
63.9
59.2
55.4
52.2
49.5
47.2
45.2
43.4
41.8
40.4
39.1
38.0
36.9
35.9
35.0

of complication
betweenthe estimatesof the probability
irradiation.
Thedifferences
uniformpartial-volume
give a measureof the clinicallyimportantdifferencesbetweenthe histograms.Incidentally,with a target
will yielda clinicallysignificant
the two-fieldarrangements
dose of 66 Gy, only the histogramrepresenting
probabilityof complication.In general,the comparisonof the dose-volumehistogramsof the sameorgan
TABLEIII
Factorsfor DerivingToleranceDoses
Organ/tissue

TDso

Bladder
Bone
Brain
Brainstem
Caudaequina
Esophagus
Eye
Femoralhead
Heart
Intestine
Kidney
Liver
Lung
Mandible
Opticnerve
Parotid
Pituitary
Rectum
Spinalcord
Stomach

72.0
70.0
64.0
64.0
57.5
66.0
60.0
62.0
41.9
55.0
29.0
35.0
22.0
77.0
65.0
72.0
54.0
75.0
50.0
55.0

0.10
0.05
0.20
0.05
0.10
0.15
0.05
0.05
0.50
0.10
0.15
0.40
0.65
0.05
0.05
0.10
0.05
0.10
0.10
0.35

0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

S-17

COMPLICATION
PROBABILITY

(D

E
E

0.6

0.4 0.2 -

0
0

10

30

20

40

50

60

Percent Target Dose

FIG.2. Cumulativedose-volumehistogramsfor the incidentalirradiationof a kidney.Differencesare
relatedto the numberof treatmentfieldsandthe typeof radiationemployed.
of tworivalplansmaynotbeas simpleastheseinitialconsiderations
maysuggestbecausethetwodistributions
mayhavedissimilarshapes(Fig.3). Thedifferencebetweenthesehistogramsis thatone histogramrepresents
a treatmentplanthattakesa smallvolumeto a higherdose whileminimizingthe dose to a largervolume
whilethe otherplan does the converse.The clinicalsignificanceof the dose distributionsrepresentedby
thesehistogramsdependsuponthe typeof tissue.Sometissueswillbe moretolerantof small-volumehighdose regionsthanothers.To estimatecomplicationprobabilities
forthe generalcase(one whichcannotbe
approximated
by a single-stephistogram),a recursivehistogram-reduction
algorithmhas beendeveloped.3
The algorithmis usedto determinean estimateof the complicationprobabilityfromthe two highestdose
stepsof the histogram,andthenthe algorithmis recursivelyusedto sumthe partialcontributionsfromthe
remainingstepsof the histogram.The measureof the complicationusedforthe histogramreductionis the
standardized
normaldeviate,whichis the numberof standarddeviationsbetweenthe mean(TD50(V))and
the dose for the dose-volumestep in question(t = (D(V) - TD5o(V))/a(V))).
Implicitin this procedureis
theassumptionthatall regionsof an organ,represented
in a singlehistogram,havethesameradiosensitivity,
andno portionof the organis morecriticalthananotherfornormalfunction.2Regionsof tissuesor organs
with differentradiosensitivities
can be handledin separatehistogramswith differenttoleranceparameters(8).
When thereis more than one normalstructureat riskfor a severecomplication,the estimatesof the
complicationscan be combinedto determinethe probabilityforany complication(9).
N

Pc=

Pc()).

(3)

i=l

Failureto controlthe tumorcan alsobe consideredas a treatmentcomplication.A probabilityof tumorcontrolalgorithm(10),whichusesthedose-volumehistogramforthetargetvolume,hasbeenimplemented.

forlocalcontrol,
forbeingcomplicationfree,Pf = 1 - P,, andtheprobability
Bycombiningtheprobabilities
localcontrol,P,fc.(11). This probability
PIc,it is possibleto estimatethe probabilityof complication-free,
is then usedto obtainan optimizedtreatmentplan.
RESULTS

There are some important deductions that can be made from the estimates of the
complication probability as represented by Eqs. (1) and (2). If the treatment plan is
3J. T. Lymanand A. B. Wolbarst,A generalmethodof assessingcomplicationsfrom dose-volume
histograms.Manuscriptsubmittedforpublication.

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

S-18

JOHN T. LYMAN

0.8 -

E
06-

0.40.2 02
0

10

20

30

40

50

60

Dose (Gray)
FIG. 3. Cumulativedose-volumehistogramsfor the incidentalirradiationof a kidney.Theseare more
thanthoseof Fig.2 andthebetterdistribution
is notobvious.Estimates
of thecomplication
complexhistograms
probabilitycan aid in the selectionof the betterplan.

designed to have a 5% complication probability, then an error which would result in

a 5% overdose will double the complication probability. Similarly, if the tolerance
dose is 5% lower than expected the complication probability will also be doubled.
Errors in the estimation of the tolerance dose can result from the estimation of the
volume irradiated. Assuming 50% of the heart was irradiatedwhen it was really 60%
will result in a 10%errorin the tolerance dose (Table II). Factors such as fractionation
scheme or adjuvant chemotherapy can also result in changes to the tolerance dose.
The algorithm used to determine the tolerance doses includes corrections for fraction
sizes other than 2 Gy (12).
The histogram-reduction algorithm gives the expected result for simple histograms
(Fig. 2). The expected result is the estimate calculated by Eqs. (1) and (2) when the
histogram is approximated by a single step. For the more general histograms
(Fig. 3), where there are no known tolerance dose data, there are no data for comparison.
The estimates in these cases appear reasonable when one considers the dose-volume
dependence of the tissue (Table III).
DISCUSSION

Individual variations in size and location of tumors can result in higher or lower
estimates of complications. If these individual variations are taken into consideration
during the treatment planning process, better (lower likelihood of complications) plans
may result. This can be considered to be a predictive assay based on patient anatomy
and the treatment plan. The result is the identification of the combination of patient
and treatment plan which have high or low likelihood for complications. This may
result in deviating from standard protocols by using a lower tumor dose for some
patients and a higher dose for other patients with similar type tumors.
This approachto optimization of treatment plans is being tested in an NCI-supported
project on the evaluation of particletreatment plans. The estimates of the complication
probabilities are also being compared with the clinical observations following heavy-

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions

PROBABILITY
COMPLICATION

S-19

Thesecomparisonswill be helpfulin establishingthe

irradiations.1'2
charged-particle
of
the
algorithmand the data in TableIII.
validity
histogram-reduction
Whiletherearea numberof factorsthatcontributeto theprobability
of complication,
it is well establishedthatdose, volume,and fractionationareimportantfactors.Since
the complicationprobabilitycan be verysensitiveto the dose and the tolerancedose,
accurateestimatesof the dose-volume distributionswithin organsand tissues are
needed.This necessitatestreatmentplanningthe entirevolume containingthe structuresof interest.From the dose-volume histogramsand the tolerancedoses, it is a
simpleprocedureto estimatethe complicationprobabilities.These estimatescan be
usedto optimizetreatmentplans.Treatmentplansarenot optimizedunlesstheyhave
takeninto accountthe individualvariationsin size, shape,and locationof the normal
criticalstructuresrelativeto the targetvolume.Comparisonsof the observedand the
expectedcomplicationprobabilitieswould resultin betterpredictionsof the complicationsarisingfromthe nonuniformirradiationof the normaltissues.
ACKNOWLEDGMENTS
Thisworkwassupportedin partby the NationalCancerInstitute,NationalInstitutesof Health,Public
HealthServiceunderAgreementNo. Y01-CM-20211andtheOfficeof HealthandEnvironmental
Research,
Officeof EnergyResearch,U. S. Departmentof EnergyunderContractDE-AC03-76SF00098.
REFERENCES
of
and preventability
and F. A. GIBBS,Preventionof radiationinjury:Predictability
1. J. R. STEWART
complicationof radiationinjury.Annu.Rev.Med.33, 385-395 (1982).
treatmentplanning:I. Delineationof anatomy.Int.
2. M. GOITEINand M. ABRAMS,Multi-dimensional
J. Radiat.Oncol.Biol.Phys.9, 777-787 (1983).
J. R. CASTRO,J. M. COLLIER,J. T. LYMAN,S. PITLUCK,
3. G. T. Y. CHEN, M. AUSTIN-SEYMOUR,
W. M. SAUNDERS,and S. R. ZINK,Dose volumehistogramsin treatmentplanningevaluationof

on Usesof Computers
carcinomaof the pancreas.In Proceedings,EighthInternational
Conference
in RadiationTherapy,pp. 264-268. IEEE,1984.
A direction for clinical radiation pathology: The tolerance dose. In
4. P. RUBINand G. W. CASARETT,
Frontiers of Radiation Therapy and Oncology (J. M. Vaeth, Ed.), Vol. 6, pp. 1-16. University Park

Press,Baltimore,1972.
5. P. RUBIN,R. A. COOPER,and T. L. PHILLIPS,
(Eds.), RadiationBiologyand RadiationPathology
Syllabus, Set R. T. : Radiation Oncology, pp. 2-7. American College of Radiology, Chicago, 1975.
6. L. COHEN,The tissue factor in radiation oncology. Int. J. Radiat.Oncol.Biol. Phys.8, 1771-1774
(1982).
7. H. R. WITHERS, H. D. THAMES, and L. J. PETERS, Dose fractionation and volume effects in normal
tissues and tumors. Cancer Treat. Symp. 1, 75-83 (1984).
C. G. ORTON,and R. A. PECK,Models in radiotherapy: Volume effects. Med.
8. T. E. SCHULTHEISS,
Phys. 10, 410-415 (1983).
L. M. CHIN, and G. K. SVENSSON,
9. A. B. WOLBARST,
Optimization of radiation therapy: Integral-

responseof a modelbiologicalsystem.Int.J. Radiat.Oncol.Biol. Phys.8, 1761-1769(1982).

10. M. GOITEIN, The utility of computed tomography in radiation therapy: An estimate of outcome. Int.

J. Radiat.Oncol.Biol.Phys.5, 1799-1807(1979).
11. L. COHEN,BiophysicalModelingin RadiationOncology.CRC Press, Boca Raton, 1983.
fornormaltissueresponses.
doserateandiso-effectrelationships
Dosefractionation,
12. G. W. BARENDSEN,
Int.J. Radiat.Oncol.Biol. Phys.8, 1981-1997(1982).

This content downloaded from 200.3.154.202 on Fri, 20 Feb 2015 13:02:23 PM

All use subject to JSTOR Terms and Conditions