Quantification of growth patterns of screen-detected lung cancers: The NELSON study

(1)

ContentslistsavailableatScienceDirect

Lung

Cancer

j o ur n a l ho me p ag e : w w w . e l s e v i e r . c o m / l o c a t e / l u n g c a n

Quantiﬁcation

of

growth

patterns

of

screen-detected

lung

cancers:

The

NELSON

study

Marjolein

A. Heuvelmans

a,b,∗

_,

_Rozemarijn

_Vliegenthart

a

_,

_Harry

_J.

_de

_Koning

c

_,

Harry

J.M.

Groen

d

_,

_Michel

_J.A.M.

_van

_Putten

e,f

_,

_Uraujh

_Yousaf-Khan

c

_,

_Carla

_Weenink

g

_,

Kristiaan

Nackaerts

h

_,

_Pim

_A.

_de

_Jong

i

_,

_Matthijs

_Oudkerk

a

a_University_of_Groningen,_University_Medical_Center_Groningen,_Center_for_Medical_Imaging_–_North_East_Netherlands,_Hanzeplein_1,₉₇₁₃_GZ_Groningen,

TheNetherlands

b_Department_of_Pulmonology,_Medisch_Spectrum_Twente,_Koningsplein_1,₇₅₁₂_KZ_Enschede,_The_Netherlands

c_Department_of_Public_Health,_Erasmus_University_Medical_Center,_P.O._Box_2040,_3000CA_Rotterdam,_The_Netherlands

d_University_of_Groningen,_University_Medical_Center_Groningen,_Department_of_Pulmonology,_Hanzeplein_1,₉₇₁₃_GZ_Groningen,_The_Netherlands

e_Department_of_Clinical_{Neurophysiology,}_MIRA,_Institute_for_Biomedical_Technology_and_Technical_Medicine,_University_of_Twente,_Enschede,_The

Netherlands

f_Department_of_Neurology_and_Clinical_{Neurophysiology,}_Medisch_Spectrum_Twente,_Koningsplein_1,₇₅₁₂_KZ_Enschede,_The_Netherlands

g_Department_of_Pulmonary_Medicine,_Kennemer_Gasthuis,_{Boerhaavelaan}_22,₂₀₃₅_RC_Haarlem,_The_Netherlands

h_Department_of_Pulmonary_Medicine,_KU_Leuven_–_University_Hospital_Leuven,_Herestraat_49,₃₀₀₀_Leuven,_Belgium

i_Department_of_Radiology,_University_Medical_Center_Utrecht,_{Heidelberglaan}_100,₃₅₈₄_CX_Utrecht,_The_Netherlands

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received15January2017

Receivedinrevisedform23February2017

Accepted25February2017 Keywords: Lungneoplasms Pulmonarynodule Growthcharts Massscreening

a

b

s

t

r

a

c

t

Objectives:Althoughexponentialgrowthisassumedforlungcancer,thishasneverbeenquantiﬁedin

vivo.Aimofthisstudywastoevaluateandquantifygrowthpatternsoflungcancersdetectedinthe

Dutch-Belgianlow-dosecomputedtomography(CT)lungcancerscreeningtrial(NELSON),inorderto

elucidatethedevelopmentandprogressionofearlylungcancer.

Materialsandmethods:Solidlungnodulesfoundat≥3CTexaminationsbeforelungcancerdiagnosis

wereincluded.Lungcancervolume(V)growthcurveswereﬁttedwithasingleexponential,expressed

asV=V1exp(t/),withttimefrombaseline(days),V1estimatedbaselinevolume(mm3),andestimated

timeconstant.TheR2_coefﬁcient_of_{determination}_was_used_to_evaluate_goodness_of_ﬁt._Overall

volume-doublingtimefortheindividuallungcancerisgivenby*log(2).

Results:Forty-sevenlungcancersin46participantswereincluded.Fortyparticipantsweremale(87.0%);

meanagewas61.7years(standarddeviation,6.2years).Mediannodulesizeatbaselinewas99.5mm3

(IQR:46.8–261.8mm3_)._Nodules_were_followed_for_a_median_of₇₇₀_days_{(inter-quartile}_range:_383–1102

days)beforelungcancerdiagnosis.Onecancer(2.1%)wasdiagnosedaftersixCTexaminations,sixcancers

(12.8%)werediagnosedafterﬁveCTs,14(29.8%)afterfourCTs,and26cancers(55.3%)afterthreeCTs.

Lungcancergrowthcouldbedescribedbyanexponentialfunctionwithexcellentgoodnessofﬁt(R2

0.98).Medianoverallvolume-doublingtimewas348days(inter-quartilerange:222–492days).

Conclusion:ThisstudybasedonCTlungcancerscreeningprovidesinvivoevidencethatgrowthof

can-ceroussmall-to-intermediatesizedlungnodulesdetectedatlow-doseCTlungcancerscreeningcanbe

describedbyanexponentialfunctionsuchasvolume-doublingtime.

BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Abbreviations: CI,conﬁdenceinterval;CT,computedtomography;IQR,inter-quartilerange;MDCT,multi-detectorcomputedtomography;NELSON,Dutch-Belgian

randomizedlungcancerscreeningtrial;V,volume(mm3_);_VDT,_{volume-doubling}_time_(days);_,_time_constant.

∗ Correspondingauthorat:UniversityofGroningen,UniversityMedicalCenterGroningen,CenterforMedicalImaging–NorthEastNetherlands,Hanzeplein1,9713GZ

Groningen,TheNetherlands.

E-mailaddresses:m.a.heuvelmans@umcg.nl(M.A.Heuvelmans),r.vliegenthart@umcg.nl(R.Vliegenthart),h.dekoning@erasmusmc.nl(H.J.deKoning),

h.j.m.groen@umcg.nl(H.J.M.Groen),m.vanputten@mst.nl(M.J.A.M.vanPutten),a.yousaf@erasmusmc.nl(U.Yousaf-Khan),weenink@kg.nl(C.Weenink),

kristiaan.nackaerts@zuleuven.be(K.Nackaerts),pimdejong@gmail.com(P.A.deJong),m.oudkerk@umcg.nl(M.Oudkerk).

http://dx.doi.org/10.1016/j.lungcan.2017.02.021

(2)

1. Introduction

Atumorstartsfromasinglecellthatapproximatelydoublesin volumewitheachcelldivision.Basedonthisprinciple,a theoreti-calexponentialmodelfortumorgrowthwasalreadyintroduced in 1956[1].It is assumed that thismodel alsoﬁts lung cancer growth,howeverthishasnotbeenproveninclinicalpractice.A wait-and-seeprincipleisnotcommonlyusedwhenlung cancer issuspected,becauseoftheaggressivenessofthedisease.Invivo informationonlungcancergrowthpatterns,fromsmallnodules barelydetectablebyimagingtechniques(∼15mm3₎_to histologi-callyprovenlungcancers,isthereforescarce.Withtheintroduction oflow-dosecomputedtomography(LDCT)lungcancerscreening, small-to-intermediatesizednodules,usuallybenign,arefoundin themajorityofscreenees[2–4].Follow-upCTsareperformedto determinenodulegrowth,inordertodifferentiatebetweenbenign andmalignantnodules.Screeningprovidesauniqueopportunity toanalyzelungcancergrowthpatternsinmoredetail.Sofar,there isonestudyinwhichgrowthdynamicsofuntreated,subclinical screen-detectedlungcancersonCTexaminationswereevaluated

[5].Theyconcludedthatmostoftheeighteenlungcancersstudied didnotshowexponentialgrowth[5].

Currently,lungcancerscreeningbyLDCTisalreadybeing imple-mentedindailypracticeintheUnitedStates.InEurope,decisions regardingtheintroductionoflungcancerscreeningwilldependon theﬁnalresultsofongoingtrials[6].AchallengeinCTlung can-cerscreeningis thehighrateoffalse-positivescreenresults.At ﬁrstdetection,nodulemanagementisbasedonnodulesize.Inthe follow-upofbothscreen-detectednodulesandnodules inciden-tallydetectedinroutineclinicalcare,accuratedetectionofnodule growth is essential to reduce false-negative and false-positive screenresults.Onewaytodeterminenodulegrowthisby calcu-latingthevolume-doublingtime(VDT).However,thisbiomarker isbasedontheassumptionthatlungcancergrowsexponentially.

InthestudybyLindelletal.,mentionedabove,growthcurves werebased ondiameter-basedestimated volumes[5]. Further-more,lungcancergrowthpatternswereonlyvisuallyevaluated andnotquantiﬁed.IntheDutch-Belgianrandomizedlungcancer screeningtrial(Dutchacronym:NELSON), nodulevolumeswere generatedsemi-automaticallybysoftware;afarmoreaccurateway todeterminenodulesize[7,8].Inaddition,forsubsequent screen-ingroundsVDTsweredeterminedbasedonnodulevolume[7,8]. Thisallowsdetailedevaluationoflungcancergrowth,including quantiﬁcationofgrowthcurves.Thepurposeofthisstudywasto evaluateandquantifygrowthpatternsoflungcancersdetectedin LDCTlungcancerscreening,inordertoelucidatethedevelopment andprogressionofearlylungcancer.

2. Materialsandmethods

2.1. Participants

The Dutch Ministry of Health approved the NELSON trial (ISRCTN63545820).Participantsprovided writteninformed con-sent.Recruitmentandselectioncriteriahavebeenpublished[9]. Participants(heavy(ex-)smokers,aged50–75years)were random-izedtonoscreening(n=7907)orscreening(n=7915)bychestLDCT atbaseline(1stround),after1year(2ndround),3years(3rdround) and5.5years(4thround)[10].

Forthissub-study,solidlungnodulesdiagnosedaslungcancer afterthreeormoresubsequentLDCTscanswithsuccessful semi-automatedvolumemeasurementswereselected.Twolungcancers wereexcludedbecauseoflackofathirdvolumemeasurementdue tofailingvolumemeasurementswithavolumetoolargefor semi-automatedmeasurementsattheﬁnal CT.Furthermore,subsolid

lungcancerswereexcluded,sincetheSyngoLungCARE©software package wasnotabletocalculatethevolumeof sub-solid nod-ules(1.9%ofallnon-calciﬁednodules).Growthcurvesofnineteen subsolidNELSONlungcancershavebeenpublishedelsewhere[33]. 2.2. Equipmentandimagereading

Sixteen-multi-detector (MD)CT scanners or, in later rounds, 64-MDCT scanners were used (Sensation-16 or Sensation-64, SiemensMedicalSolutions,Forchheim,Germany,orMx8000IDT or Brilliance 16P or Brilliance 64, Philips Healthcare, Best, The Netherlands).Scanningparametershavebeenpublished,andwere equalforallscreenings[11,12].

Lung nodule evaluation was performed with software for semi-automatedvolumemeasurements(LungCARE©[Somaris/5 VA70C-W,SiemensMedicalSolutions])[11].Intheﬁrsttwo screen-ingrounds,CTimageswerereadtwice[11,12];ifmeasuredvolume differedbetweentheindependentreaders,secondreader’s mea-surementswereused.Inthenextscreeningrounds,oneradiologist withatleast6yearsexperiencereadtheexaminations.

Nodulemarginwasclassiﬁedassmooth,lobulated,spiculated orirregular,basedonthethree-dimensionalnodulesegmentation

[13–15].

2.3. Nodulemanagementprotocol

The nodule management protocol was previously described

[11].Screeningscouldleadtothreeoutcomes:negativeresult (vol-ume<50mm3_;_implication:_next_screening_round),_{indeterminate} result(volume50–500mm3_;_short-term_follow-up_CT_after_6–12 weeks)orpositive result(volume>500mm3_;_referral_to pulmo-nologist).For previously detectednodules, atleast 25%volume increase led toVDT assessment. Nodules with VDT <400 days resultedinapositiveresult[11].Positivescreeneeswerereferred toapulmonologistfordiagnosticworkup.Theyreceivedusualcare accordingto(inter)-nationalguidelines[16–18].NELSON’s chief pathologistreassessed histologicalspecimens ofpositive scree-nees.

2.4. Growthcurvesandstatisticalanalyses

Lungcancervolume(V)growthcurveswereﬁttedasfunction oftimet,assuminganexponentialgrowthpattern,using

V=V1·exp

_t

(1) withttimefrombaseline,V1thevolumeatbaselineinmm3and thetimeconstant(days).Curveswerefittedusingaleastsquare estimationprocedure,returningestimatesforV1and.As mea-sureforthegoodnessoffit,R2_coefficient_{determination}_was_used, where a perfectfit resultsin R2₌_1. _Normalized_growth_curves, includingalllungcancers,werecreatedbyplottingnormalized vol-ume(V/V1)onalogarithmicy-axisasfunctionofnormalizedtime, using

t∗= t

. (2)

OverallVDTperlungcancercasewascalculatedfromtheestimated timeconstantusing

VDT=·log(2) (3)

DifferenceinoverallVDTbetweennodulesdetectedatdifferent numbersofscreening CTexaminationsbeforediagnosisoflung cancerwastestedbytheKruskal–Wallistest.P≤0.05was consid-eredtoindicatestatisticalsigniﬁcance.Analyseswereperformed

(3)

Fig.1.RadarplotshowingindividualR2_-deviation_for_all₄₇_lung_cancers.

Fig.2.CTexaminationswithsemi-automatedvolumemeasurementsofalungnodulediagnosedasbeingmalignantafterfoursubsequentCTexaminations.Thisnodule

grewfrom17.0mm3_at_baseline_to_88.3_mm3_at_time_of_referral_after_follow-up_examination_in_the_second_screening_round,_1.5-years_after_baseline._Growth_curve_of_this

noduleisshowninFig.3b.

usingSPSS20.0(SPSS,Chicago,Ill,USA),andOctave(www.octave. org).

3. Results

3.1. Participants

Intotal,46participantswith47screen-detectedlungcancers basedonasolidlungnodulefoundat≥3LDCTsbeforereferraltoa pulmonologist,wereincluded.Meanageoftheseparticipantswas 61.7years(standarddeviation,6.2years),40weremale(87.0%). Twenty-sevenparticipants(58.7%) werecurrentsmokers;mean smokedpack-years was47.6(21.1).Nodules werefollowed for amedianof770days(IQR: 383–1102days),beforelungcancer wasdiagnosed.Mediannodulesizeatbaselinewas99.5mm3_(IQR: 46.8–261.8mm3_)._One_cancer_(2.1%)_was_referred_for_work-up_and diagnosedaftersixCTexaminations,six(12.8%)afterﬁveCTs,14 (29.8%)afterfourCTs,and26cancers(55.3%)werediagnosedafter threeCTexaminations.

3.2. Cancercharacteristics

LungcancercharacteristicsarepresentedinTable1.Attimeof diagnosis,themajorityoflungcancerswasstageI(35/47,74.5%). Mostcancerswereadenocarcinoma(38/47,80.9%).Furthermore, foursquamous cellcarcinomas (8.5%), onelargecellcarcinoma (2.1%),andoneatypicalcarcinoid(2.1%)wasfound.Twocancers (4.2%)were diagnosedasnon-small celllung cancernot other-wisespeciﬁed.Onenodule(2.1%)wastreatedasalungcancerby stereotacticradiotherapy,withoutﬁnalhistologicaldiagnosis.

3.3. Growthpatterns

GoodfittoexponentialgrowthwasconfirmedbythehighR2 coefficientofdeterminationfortheindividualcancergrowthcurves (median0.98;IQR:0.94–0.99).AnoverviewoftheR2_coefficient_of determinationfortheindividuallungcancersisshowninFig.1. R2 _of _pulmonary _nodules _found _at _≥4 _CT _examinations _before lungcancerdiagnosisdidnotdiffersignificantlyfromR2 _of

(4)

nod-Fig.3. Examplesoflungcancergrowthpattern.(a)Growthcurveofalungnodule

thatwasdetectedonsixsubsequentCTexaminationsbeforelungcancerdiagnosis

wasmade.Thenodulegrewfrom264.8mm3_at_baseline_to_1038.8_mm3_at_time_of

referralafterthefourthscreeninground,5.5-yearsafterbaseline.Theexponential

ﬁt(usingEq. (1)),is nearly perfect(R2₌_0.97),_with _estimated_initial _volume

V1=276.5mm3andtimeconstant=1484days,resultinginanoverall

volume-doublingtime(VDT)of1027days.(b)Growthcurveofalungnodulediagnosed

asbeingmalignantafterfoursubsequentCTexaminations.Thisnodulegrewfrom

17.0mm3_at_baseline_to_88.3_mm3_at_time_of_referral_after_follow-up_examination_in

thesecondscreeninground,1.5-yearsafterbaseline.Theﬁttedexponential(with

Table1

Characteristicsofsolidlungcancersdiagnosedafter≥3CTexaminations.

Characteristics Lungcancernodule(n=47) Histologicaltype(n(%))

Adenocarcinoma 38(80.9)

Squamouscellcarcinoma 4(8.5) Largecellcarcinoma 1(2.1)

Carcinoid 1(2.1) NSCLC,notspeciﬁed 2(4.2) Nohistologyobtained 1(2.1) Cancerstage(n(%)) IA 33(70.2) IB 2(4.3) IIA 3(6.4) IIB 1(2.1) IIIA 4(8.5) IIIB 1(2.1) IV 3(6.4)

Follow-uptimeafterbaseline(days)

Median(IQR) 770(383–1102) Margin(n(%)) Smooth 5(10.6) Lobulated 29(61.7) Spiculated 7(14.9) Irregular 6(12.8)

Note:Unlessotherwiseindicated,dataarenumbersofnodules,withpercentagesin parentheses.IQR,inter-quartilerange.

ulesfoundatthreeCTexaminations(median0.97[IQR:0.91–0.99]

versus0.99[IQR:0.96–1.00],respectively[p=NS]).

InFigs.2and3,examplesof,respectively,CTimagesandgrowth patterns of lung cancers are shown. Not all cancers showed a growthpatternwithfastgrowthimmediatelyfromﬁrstdetection; ﬁvenodulesshowedanalmostconstant(small)volumeforatleast 500daysbeforegrowthexpansionanddiagnosisoflungcancer,see

Fig.3c.Still,growthpatternsofthesenodulescouldbedescribedby theexponentialfunctionwithanexcellentﬁt(median1.00[IQR: 0.98–1.00]).Afternormalizationofvolumemeasurementsandof timefrombaselinefortheindividuallungcancers,lineargrowthon alogarithmicscalewasfound,indicativeofanexponentialgrowth pattern(Fig.4).

3.4. Overallvolume-doublingtimes

Median overall VDT from baseline until last screen evalua-tionbeforecancerdiagnosiswas348days(IQR:222–492);range 80–4271days.WhencomparingoverallVDTofnodulesfollowed foradifferentnumberofCTexaminationsbeforelungcancer diag-nosis,VDT turnedout tobelowerin participantswhoreceived lessCTs(median270days[IQR:182–357days]incaseofthree CTs,median368days[IQR:305–554days]incaseoffourCTs,and median960days[IQR:462–1039days]incaseofﬁveorsixCTs [p<0.01]).Anoverviewofthenumberoflungcancersperoverall VDTisshowninFig.5.

3.5. Non-ﬁttinggrowthcurves

Althoughmostlungcancersshowedanexcellentﬁttoan expo-nentialcurve,thereweresomeexceptions.TwocaseswithR2_<_0.5 werevisuallyre-evaluated.Inonecancer,semi-automatedvolume measurementswerenot accuratesincethis nodulewaspleural attached.Inthis case,thesoftwarewasnotabletodifferentiate

R2₌_0.99)_returns_V

1=20.2mm3and=334dayswithoverallVDT=232days.(c)

Exampleofacancershowingaperiodofconstantsmallvolumebeforegrowth

expansion.Shownisthegrowthcurveofalungnodulethatwasdetectedonﬁve

sub-sequentCTexaminationsbeforelungcancerdiagnosiswasmade.Thenodulegrew

from65mm3_at_baseline_to₅₀₄₂_mm3_at_time_of_referral_after_the_fourth_screening

(5)

Fig.4.Graphofnormalizedgrowthcurvesoftheindividuallungcancers.The

y-axisrepresentsnormalizednodulevolume(V/V1)onalogarithmicscale.Thex-axis

representsnormalizedtimet*=t/.

Fig.5.Overallvolume-doublingtime(VDT)for46lungcancers.Onecasewith

outlierVDTof4271dayswasnotdepictedinthisﬁgure.

pleurafromnoduleandunderestimatednodulevolume,andthe participantwasreferredbasedonvisualnodulegrowth.Thesecond casecomprisedaparticipantwithadoubletumor.Theparticipant receivedapositivetestresultbasedonalarger,fast-growing nod-ule.Attimeofsurgery,however,alsoasmallernoduleinanother lobewasresected.Thisnoduleturnedouttobelungcanceraswell.

4. Discussion

Weshowthatgrowthoflungcancers,detectedinaLDCTlung cancerscreeningtrial,usuallycanbedescribedbyanexponential growthfunction.Informationongrowthpatternsoflungcancer isofgreatimportance,sinceaftertheintroductionandcontinual improvementofCT,andthemorerecentintroductionoflung can-cerscreening byLDCT, alargenumber ofnodules withsmaller dimensionsisdetected,mostbeingbenign[2–4].Forthose small-to-intermediatesizedpulmonarynodules,thedeﬁnitivediagnosis stillremainsa dilemmaforradiologists.Asmorphologicnodule evaluationaloneisnotsufﬁcient[19],follow-upisnecessary.For intermediate-sized nodules, the decision for (invasive) workup

usuallydependsongrowthrate.Themalignantpotentialofthese nodulesisoftensuggestedbyrapidgrowthrateintermsofashort VDT[4,20].TheuseofVDT,however,isbasedontheassumptionof anexponentialgrowthpatternoflungcancers.

Thismulti-centerstudyisoneoftheﬁrsttoactuallyquantify andconﬁrm,notonlybyvisualanalysis,thatgrowthinearlylung cancerscanbedescribedbyusing anexponentialfunction,and indicatesthatlungcancergrowthcanbemonitoredbytheVDTas imagingbiomarker.Itprovidesuniqueinsightinthenatural behav-ioroflungcancer;atypeofmalignancythat,untreated,usually cannotbefollowedoveralongertimeperiodinclinicalcare.

ThemedianoverallVDToflungcancersinthisstudywas348 days(IQR: 222–492days),and comparabletoVDTsreportedin other lung cancerscreening trials [21–23]. Since at least three monitoringpoints arenecessary for growthpattern evaluation, lungcancersreferredaftertheﬁrstshort-termfollow-upCTwere notincludedinthisanalysis.Thelatternodulesusuallyshowfast growthratesbelow232days[24].

Ithasbeensuggestedthat lungcancersgrowaccordingtoa Gompartziangrowthmodel,whichmeansthatgrowthrateslowly decreasestoreachacertainplateau,afteraperiodofexponential growth[25,26].Inthecurrentstudy,inwhichthemajorityoflung cancerswasearlystage,wefoundthatcancersshowedexponential growththroughouttheobservationalperiod.Adecreaseingrowth ratewasnotobservedpossiblybecausetumorsweretoosmallto reachthatpointatdiagnosisandparticipantsweretreatedafter lungcancerdiagnosiswasmade.

Lindell et al.[5] werethe first toplot growth curves of 18 lung cancers diagnosed after at least four serial CT examina-tions.Itwasconcludedthatlungcancergrowthwasnotlimited toan exponential pattern,and thus questioned theapplication ofVDT aspartof nodulemanagementalgorithms inLDCTlung cancer screening. That study, however, had some limitations. First,themajorityofstudiedlungcancers(n=11;61%)was sub-solid. Volume for all nodules was calculated based on nodule diameter (V=1/6**[ab2_], _with _a _the _longest _and _b _the per-pendiculardiameter).Especiallyinsub-solidnodules,butalsoin solidnodules,diametermeasurementsarelessaccuratethan vol-umemeasurements[7,8].Furthermore,thestudyusedrelatively thickslice-thickness(3.75–5mm)andlargereconstruction inter-vals(5mm)thatmayhavecontributed toeven moreimprecise diametermeasurements,especiallyforsub-centimeternodules.In theNELSONstudy,slicethicknessof1.0mmwithreconstruction intervalof0.7mmwasusedtoprovideisometricvoxelsforaccurate semi-automatedvolume measurements.These semi-automated measurementswerefoundtobehighlyreproducibleforthelarge majorityofsolidnodules,alsoinphantomstudies[27–29],with highreaderagreementandvolumedifferences>15%betweentwo readersinonly4%ofsolidnodules[27–29].Besides,inLindell’s studylungcancergrowthcurveswereplottedandvisually evalu-ated,butgrowthpatternswerenotquantified.Theyonlyincluded lungcancersdiagnosedafter≥4CTexaminations,sincetheystated thatfourwastheminimalnumberofmonitoringpointsrequired forgeneratinggrowthcurves[5].Wehaveincludedcancers diag-nosedafterthreeCTsaswell.Wefoundthattheexponentialfitin termsoftheR2_coefficient_of_{determination}_of_lung_cancers diag-nosedafter≥4CTswascomparabletolungcancersdiagnosedafter threeCTs.

Asecondstudyevaluatedgrowthcurvesof thirteenprimary lungcancerswithatleastthreeserialCTexaminations[30].Nodule volumeswerederivedsemi-automatically.Analysisoflungcancer growthcurveswaslimited,andrestrictedtoaplotofthegrowth curvesonalogscale,onlyanalyzedvisually.Growthratesappeared approximatelyconstantonthelogscale,consistentwith exponen-tialgrowth.

(6)

Inourstudy,ﬁveof47lungcancersshowedanalmostconstant (small)volumeforatleast500daysbeforegrowthexpansionand diagnosisoflungcancer.Itisknownthatthecarcinogenicprocess leadingtolungcancerevolvesslowlydependingonthe accumula-tionrateofDNAaberrationsintheancestorcells,forover20years

[31].Inascreeningsetting,incidentaldetectionofaslowergrowing cancerismorelikelythanitisingeneralpractice.Anothercause fortheperiodofconstantvolumemaybeaproliferationratein balancewiththeapoptoticrate.

Inlungcancerscreeningprogramsandinclinicalpractice, nod-ulemanagementisnotonlybasedonnodulesizeatfirstdetection, but also on change in size in case of a prevalence nodule. In mostdiameter-basednodulemanagementprotocols,includingthe recentlypublishedLung-RADS guideline [32],nodule growthis definedasafixedincreaseindiameter,regardlessofthetime inter-valbetweentwosubsequentscreenings.Ourstudysuggeststhat thislinearmethodcannotaccuratelyrepresenttherateofnodule growth.Alinearmethodmightleadtofalse-negativeclassification ofsmall,fast-growingcancersatshort-termfollow-up.

Themajorlimitationofthisstudyisthefactthatonlygrowth patternsofrelativelyslow-growinglungcancers(aboutone-fifth oftotalscreen-detectedcancers)couldbeevaluated,sincemore aggressivelungcancersdidnothavefollow-upofatleastthreeCT examinationsbutwereearlierreferredtoapulmonologist,which maylimitthegeneralizabilityofourresults.Comparedtothe over-allgroupofscreen-detectedcancersinthefirsttothirdNELSON round,thisstudycomprisedmoreadenocarcinomas(80.9%versus 51.2%),knownasarelativelyslowergrowingtypeoflungcancer, whereasnosmallcelllungcancers(4.8%oftotalcancers),usually fast-growingandadvancedstaged,werefollowedbyatleastthree CTs.Notwithstanding,thepercentagestageIcancersinthisstudy (74.5%)wascomparabletothepercentagestageIcancersinthe overallgroupofcancersinthefirsttothirdscreeninground(70.9%). 4.1. Conclusion

ThisstudybasedonCTlungcancerscreeningprovidesinvivo evidence that growth of cancerous small-to-intermediate sized lung nodules detected at LDCT lung cancer screening can be described by an exponentialfunction suchas volume-doubling time.

Conﬂictsofinterest

MAH,RV,HJMG,MJAMvP,UY-K,CW,KN,PAdJ,MOhavenothing todisclose.HJdKreported:‘HealthTechnologyAssessmentforCT LungCancerScreeninginCanada’.CancerCareOntario,Dr.Paszat. Grant.HJdKtookpartina 1-dayadvisorymeeting on biomark-ersorganizedbyM.D.Anderson/HealthSciencesduringthe16th WorldConferenceonLungCancer.HJdKreceivedagrantfromthe UniversityofZurichtoassessthecost-effectivenessofcomputed tomographiclungcancerscreeninginSwitzerland.

Acknowledgements

TheauthorsthankthesystemcontrollersRFaberandFJP San-tegoets,and thesecretaryMQuak(all from thedepartmentof PublicHealth,ErasmusUniversityMedicalCenter)fortheir contri-butionandmaintenanceofthedatabase.Furthermore,theauthors thankRZiengs(UniversityMedicalCenterGroningen)andSvan Amelsvoort-vanderVorst(UniversityMedicalCenterUtrecht).

TheNELSONtrialissupportedby:“ZorgOnderzoek Nederland-Medische Wetenschappen” (ZonMw), “KWF Kankerbestrijding”, and“StichtingCentraalFondsReservesvanVoormaligVrijwillige Ziekenfondsverzekeringen” (RvvZ). Roche diagnostics provided

a grant for the performance of proteomics-research. Siemens Germanyprovided4digitalworkstationsandLungCARE®forthe performanceof3D-measurements.

References

[1]V.P.Collins,R.K.Loefﬂer,H.Tivey,Observationsongrowthratesofhuman tumors,Am.J.Roentgenol.RadiumTher.Nucl.Med.76(1956)988–1000.

[2]S.J.Swensen,J.R.Jett,J.A.Sloan,etal.,Screeningforlungcancerwithlow-dose spiralcomputedtomography,Am.J.Respir.Crit.CareMed.165(2002) 508–513.

[3]C.I.Henschke,D.F.Yankelevitz,D.P.Naidich,etal.,CTscreeningforlung cancer:suspiciousnessofnodulesaccordingtosizeonbaselinescans, Radiology231(2004)164–168.

[4]N.Horeweg,J.vanRosmalen,M.A.Heuvelmans,etal.,Lungcancerprobability inpatientswithCT-detectedpulmonarynodules:aprespeciﬁedanalysisof datafromtheNELSONtrialoflow-doseCTscreening,LancetOncol.15(2014) 1332–1341.

[5]R.M.Lindell,T.E.Hartman,S.J.Swensen,J.R.Jett,D.E.Midthun,J.N.Mandrekar, 5-Yearlungcancerscreeningexperience:growthcurvesof18lungcancers comparedtohistologictype,CTattenuation,stage,survival,andsize,Chest 136(2009)1586–1595.

[6]M.A.Heuvelmans,R.Vliegenthart,M.Oudkerk,ContributionsoftheEuropean Trials(EuropeanRandomizedScreeningGroup)incomputedtomography lungcancerscreening,J.Thorac.Imaging30(2015)101–107.

[7]D.F.Yankelevitz,A.P.Reeves,W.J.Kostis,B.Zhao,C.I.Henschke,Small pulmonarynodules:volumetricallydeterminedgrowthratesbasedonCT evaluation,Radiology217(2000)251–256.

[8]M.P.Revel,A.Bissery,M.Bienvenu,L.Aycard,C.Lefort,G.Frija,Are two-dimensionalCTmeasurementsofsmallnoncalciﬁedpulmonarynodules reliable?Radiology231(2004)453–458.

[9]C.A.vanIersel,H.J.deKoning,G.Draisma,etal.,Risk-basedselectionfromthe generalpopulationinascreeningtrial:selectioncriteria,recruitmentand powerfortheDutch-Belgianrandomisedlungcancermulti-sliceCTscreening trial(NELSON),Int.J.Cancer120(2007)868–874.

[10]N.Horeweg,C.M.vanderAalst,R.Vliegenthart,etal.,Volumetriccomputer tomographyscreeningforlungcancer:threeroundsoftheNELSONtrial,Eur. Respir.J.42(2013)1659–1667.

[11]D.M.Xu,H.Gietema,H.deKoning,etal.,Nodulemanagementprotocolofthe NELSONrandomisedlungcancerscreeningtrial,LungCancer54(2006) 177–184.

[12]R.J.vanKlaveren,M.Oudkerk,M.Prokop,etal.,Managementoflungnodules detectedbyvolumeCTscanning,N.Engl.J.Med.361(2009)2221–2229.

[13]S.Takashima,S.Sone,F.Li,Y.Maruyama,M.Hasegawa,M.Kadoya, Indeterminatesolitarypulmonarynodulesrevealedatpopulation-basedCT screeningofthelung:usingﬁrstfollow-updiagnosticCTtodifferentiate benignandmalignantlesions,AJRAm.J.Roentgenol.180(2003)1255–1263.

[14]D.M.Xu,H.J.vanderZaag-Loonen,M.Oudkerk,etal.,Smoothorattached solidindeterminatenodulesdetectedatbaselineCTscreeningintheNELSON study:cancerriskduring1yearoffollow-up,Radiology250(2009)264–272.

[15]J.W.Gurney,D.M.Lyddon,J.A.McKay,Determiningthelikelihoodof malignancyinsolitarypulmonarynoduleswithBayesiananalysis.PartII. Application,Radiology186(1993)415–422.

[16]J.P.vanMeerbeeck,C.C.Koning,V.C.Tjan-Heijnen,A.G.Boekema,C.J. Kaandorp,J.S.Burgers,Guidelineon‘non-smallcelllungcarcinoma;staging andtreatment’,Ned.Tijdschr.Geneeskd.149(2005)72–77.

[17]P.Goldstraw,J.Crowley,K.Chansky,etal.,TheIASLCLungCancerStaging Project:proposalsfortherevisionoftheTNMstagegroupingsinthe forthcoming(seventh)editionoftheTNMClassiﬁcationofmalignant tumours,J.Thorac.Oncol.2(2007)706–714.

[18]W.D.Travis,E.Brambilla,H.K.Muller-Hermelink,C.C.Harris,WorldHealth OrganizationClassiﬁcationofTumours,PathologyandGenetics:Tumoursof theLung,Pleura,ThymusandHeart,IARCPress,Lyon,France,2004.

[19]D.M.Xu,R.J.vanKlaveren,G.H.deBock,etal.,Limitedvalueofshape,margin andCTdensityinthediscriminationbetweenbenignandmalignantscreen detectedsolidpulmonarynodulesoftheNELSONtrial,Eur.J.Radiol.68 (2008)347–352.

[20]M.P.Revel,A.Merlin,S.Peyrard,etal.,Softwarevolumetricevaluationof doublingtimesfordifferentiatingbenignversusmalignantpulmonary nodules,AJRAm.J.Roentgenol.187(2006)135–142.

[21]D.O.Wilson,A.Ryan,C.Fuhrman,etal.,DoublingtimesandCT

screen-detectedlungcancersinthePittsburghLungScreeningStudy,Am.J. Respir.Crit.CareMed.185(2012)85–89.

[22]C.I.Henschke,D.F.Yankelevitz,R.Yip,etal.,Lungcancersdiagnosedatannual CTscreening:volumedoublingtimes,Radiology263(2012)578–583.

[23]M.Hasegawa,S.Sone,S.Takashima,etal.,Growthrateofsmalllungcancers detectedonmassCTscreening,Br.J.Radiol.73(2000)1252–1259.

[24]M.A.Heuvelmans,M.Oudkerk,G.H.deBock,etal.,Optimisationof volume-doublingtimecutoffforfast-growinglungnodulesinCTlungcancer screeningreducesfalse-positivereferrals,Eur.Radiol.23(2013)1836–1845.

[25]A.K.Laird,Dynamicsoftumorgrowth,Br.J.Cancer13(1964)490–502.

[26]X.L.Xu,ThebiologicalfoundationoftheGompertzmodel,Int.J.Biomed. Comput.20(1987)35–39.

(7)

[27]Y.Wang,R.J.vanKlaveren,H.J.vanderZaag-Loonen,etal.,Effectofnodule characteristicsonvariabilityofsemiautomatedvolumemeasurementsin pulmonarynodulesdetectedinalungcancerscreeningprogram,Radiology 248(2008)625–631.

[28]X.Xie,M.J.Willemink,P.A.deJong,etal.,Smallirregularpulmonarynodules inlow-doseCT:observerdetectionsensitivityandvolumetryaccuracy,AJR Am.J.Roentgenol.202(2014)W202–W209.

[29]X.Xie,M.J.Willemink,Y.Zhao,etal.,Inter-andintrascannervariabilityof pulmonarynodulevolumetryonlow-dose64-rowCT:ananthropomorphic phantomstudy,Br.J.Radiol.86(2013)20130160.

[30]L.E.Quint,J.Cheng,M.Schipper,A.C.Chang,G.Kalemkerian,Lunglesion doublingtimes:valuesandvariabilitybasedonmethodofvolume determination,Clin.Radiol.63(2008)41–48.

[31]E.C.deBruin,N.McGranahan,R.Mitter,etal.,Spatialandtemporaldiversity ingenomicinstabilityprocessesdeﬁneslungcancerevolution,Science346 (2014)251–256.

[32]AmericanCollegeofRadiology,LungCTScreeningReportingandDataSystem

(Lung-RADS),2015,Julyhttps://www.acr.org/Quality-Safety/Resources/

LungRADS.

[33]E.T.Scholten,P.A.deJong,B.deHoop,etal.,Towardsaclosecomputed tomographymonitoringapproachforscreendetectedsubsolidpulmonary nodules?Eur.Respir.J.45(2015)765–773.