& prevention risk factors for lymphedema after...

ReseSe

Risk

SandrHeathKevin

Abst

Intro

Ltion echanilymphtheraptors fbreasand tis notbiopsthe ax

Author2Divisioof RadiaSchooland 5Br

CorresDrive, U610-66

doi: 10

©2010

Cance2734

D

Published OnlineFirst October 26, 2010; DOI: 10.1158/1055-9965.EPI-09-1245

Cancer
Epidemiology,
iomarkersrevention

arch Articlee commentary p. 2715 B

& P

Factors for Lymphedema after Breast Cancer Treatment

a A. Norman1,4, A. Russell Localio1,4, Michael J. Kallan1, Anita L. Weber1,
er A. Simoes Torpey1, Sheryl L. Potashnik1, Linda T. Miller5, R. Fox2, Angela DeMichele1,2, and Lawrence J. Solin3
ractBac

breastMe

diagnoobtainnaire.lymphing cli

Resaxillarlymphchest w0.86 (9cyclinstageresultefor SL

Con

y (SLNBilla is bec

s' Affiliationn of HematotionOncoloofMedicineeast Cance

ponding Auniversity o

4-2350; Fax

.1158/1055-

American A

r Epidemio

ownload

kground: As cancer treatments evolve, it is important to reevaluate their effect on lymphedema risk incancer survivors.thods: Apopulation-based random sample of 631women frommetropolitan Philadelphia, Pennsylvania,sed with incident breast cancer in 1999 to 2001, was followed for 5 years. Risk factor information wased by questionnaire and medical record review. Lymphedema was assessed with a validated question-Using Cox proportional hazards models, we estimated the relative incidence rates [hazard ratios (HR)] ofedema with standard adjusted multivariable analyses ignoring interactions, followed by models includ-nically plausible treatment interactions.ults: Compared with no lymph node surgery, adjusted HRs for lymphedema were increased followingy lymph node dissection [ALND; HR, 2.61; 95% confidence interval (95% CI), 1.77-3.84] but not sentinelnode biopsy (SLNB; HR, 1.04; 95% CI, 0.58-1.88). Risk was not increased following irradiation [breast/all only: HR, 1.18 (95% CI, 0.80-1.73); breast/chest wall plus supraclavicular field (+/− full axilla): HR,5% CI, 0.48-1.54)]. Eighty-one percent of chemotherapy was anthracycline based. The HR for anthra-e chemotherapy versus no chemotherapy was 1.46 (95% CI, 1.04-2.04), persisting after stratifying onat diagnosis or number of positive nodes. Treatment combinations involving ALND or chemotherapyd in approximately 4- to 5-fold increases in HRs for lymphedema [e.g., HR of 4.16 (95% CI, 1.32-12.45)NB/chemotherapy/no radiation] compared with no treatment.clusion: With standard multivariable analyses, ALND and chemotherapy increased lymphedema riskas radiation therapy and SLNB did not. However, risk varied by combinations of exposures.
where
Impact: Treatment patterns should be considered when counseling and monitoring patients for lymphedema.Cancer Epidemiol Biomarkers Prev; 19(11); 2734–46. ©2010 AACR.

reviewALNALNDrithm(17, 2rence,tain cchemALNDCle

conce

duction

ymphedema is a common and debilitating condi-xperienced by breast cancer survivors (1-6). Me-sms and risk factors remain unclear. Axillarynode dissection (ALND) and axillary radiation

y have been cited as the most important risk fac-or lymphedema (7-15). However, approaches tot cancer diagnosis and treatment have evolved,he effect of these changes on risk of lymphedemaknown. For example, using sentinel lymph node

) as a first-line approach to evaluatingoming commonplace (16), but as recently

treatmlympaccomconcelowinadjuvirradidecisifollowreliantreatmgenei

s: 1Center for Clinical Epidemiology and Biostatistics,logy/Oncology, Department of Medicine, 3Departmentgy, and 4Department of Biostatistics andEpidemiology,, University of Pennsylvania, Philadelphia, Pennsylvania;r Physical Therapy Center, Cherry Hill, New Jersey

thor: Sandra A. Norman, 8 Blockley Hall, 423 Guardianf Pennsylvania, Philadelphia, PA 19104-6021. Phone:: 484-840-5394. E-mail: [email protected]

9965.EPI-09-1245

ssociation for Cancer Research.

l Biomarkers Prev; 19(11) November 2010

on June 1, 2018. © 201cebp.aacrjournals.org ed from

ed (17) and reported by others (18-20), whetherD should follow a positive SLNB (completion) is being actively investigated. Although algo-

s have been developed to assist decision making1), there is no consensus (17). To minimize recur-completion ALND is commonly used, and in cer-

ircumstances, axillary radiation therapy or systemicotherapy is recommended in place of completion(17).

arly, preventing cancer recurrence is the primaryrn of treatment. However, given the multiple cancerent options, the potential for such morbidities ashedema associated with axillary evaluation andpanying treatment should be considered. One suchrn is whether the lower risk of lymphedema fol-g SLNB persists after nodal irradiation or afterant chemotherapy. Older studies of axillary nodalation following ALND might not be pertinent toons about lymphedema risk when axillary radiations SLNB. Second, current practice, with greaterce on computerized tomography for radiation
ent planning, can improve radiation dose homo-ty, decrease toxicity, and avoid irradiating lymph
0 American Association for Cancer Research.

http://cebp.aacrjournals.org/

nodeunlikely anttaxan(7-14,on treindepiningsequecruciaWe

lymphdemacombpatienassociof subadditi

Mate

StudyMe

(29).Philadfirst dcanceAfterpotenwhichnewlyin theWe

patienages eobtainwrittesampfor up

DataTrai

nairecondipresenwas inmatelysocioddiagnintervdata o

VariaLym

ferencly use(31, 3based

arm mfor thmakinveloppresenseverpresesize oexpermildSpe

vieweof brerightwas rratelya locanotingage, w(handyou anoticegers’;summ0 to 9periodRes

informwhethmonteachallow0 to 9Bas

finedwas >the slymphdemadifferscoreCan

ing toativeor hogorizewall(inclubreas(incluregimanthrtherapon ancombfollow

Risk Factors for Lymphedema

www.a

D


structures when not indicated (22-26). Moreover,older studies, multiagent chemotherapy, especial-hracycline-based chemotherapy with or withoute, is common (27). According to recent reviews28), studies rarely include sufficient informationatment and patient characteristics to assess theendent and joint contributions of each. Thus, exam-currently used treatments, as well as the entirence of treatments, with respect to lymphedema isl to current oncology practice.conducted a population-based prospective study ofedema to examine, first, the association of lymphe-with breast cancer treatments individually or inination with other treatments over time, and second,t characteristics identifiable at cancer diagnosisated with increased risk of lymphedema, regardlesssequent cancer treatment, which could promptonal counseling and monitoring for lymphedema.

rials and Methods

design and study populationthods for this study have been described previouslyBriefly, eligible patients were female residents ofelphia and Delaware Counties in Pennsylvaniaiagnosed with histologically confirmed breastr between May 1, 1999 and September 30, 2001.obtaining Institutional Review Board approvals,tial participants were identified from 30 hospitals,diagnosed or treated approximately 95% of alldiagnosed primary breast cancer patients residingse two counties during the study years (30).selected an age-stratified random sample of 33% ofts ages 50 to 79 years and up to 100% of womenither <50 or ≥80 years. Physician permission wased before patient contact; all participants providedn informed consent. Women identified in thisle who were enrolled were followed prospectivelyto 5 years.

collectionned interviewers administered a structured question-on demographics, lifestyle, access to care, medicaltions, presence and timing of cancer treatments, andce and degree of lymphedema. The first interview-person, followed by telephone interviews approxi-7 to 9months apart. Except for some questions aboutemographic and preexisting conditions at cancerosis, all questions were repeated at each follow-upiew. Medical record reviews provided supplementaryn staging and specifics of cancer treatments.

ble definitionsphedema. Historically, differences in the circum-e or volume of the limbs have been most common-d to assess the presence and degree of lymphedema
2), despite the lack of consistent measurement-criteria for diagnosis (3, 8, 14, 28, 31-34). However,
Staaccord

Caacrjournals.org

on June 1, 2018. © 201cebp.aacrjournals.org ownloaded from

easurements at each follow-up were not feasibleis study, given the large cohort and the expense ofg multiple measurements over time. Thus, we de-ed a questionnaire and scoring system to assess thece and degree of lymphedema (mild or moderate/e) and the number of months lymphedema wasnt using the patient's perceived differences in thef her hands and arms, which we validated againstt clinical therapists' measurement-based criteria forand moderate/severe lymphedema (28).cifically, at the first in-person interview, inter-rs asked respondents whether, between the dateast cancer diagnosis and the interview date, theirand left hands seemed to differ in size. The questionepeated for the lower arms and upper arms sepa-. Patients who reported observing no difference attion were assigned a degree score of 0. For womensize differences, the interviewer asked, “On aver-ould you say that the difference in size of yours/lower arms/upper arms) was ‘1: very slight;re the only person who would notice this’; ‘2:able to people who know you well but not to stran-or ‘3: very noticeable’?” The degree score wased over the three locations and could range from. Subsequent telephone interviews covered the timeback to the previous interview.

pondents who reported a size difference providedation on the month and year it was first noticed,er they still noticed the difference, and if not, theh and year it returned to being the same size forpart (hand/lower arm/upper arm) separately,ing us to assign a lymphedema degree score fromto each month of follow-up.ed on the validation study, lymphedema was de-as present in any month in which the degree score0 and the limb on the side of surgery was larger. Ifcore was ≥4, the patient had moderate/severeedema; otherwise, the diagnosis was mild lymphe-(sum from 1 to 3). A score of ≥4 required size

ences at two or more locations because the largestpossible at any one location was 3 (28).cer treatments. Treatments were classified accord-the best information available, whether from oper-reports, flow sheets, physician correspondence,spital tumor registries. Radiation therapy was cate-d into the largest volume reported: breast/chestonly; breast/chest wall plus supraclavicular fieldding the apex of axilla, but not the full axilla); ort/chest wall, supraclavicular field, and full axillading a posterior axillary boost field). Chemotherapyens were recorded as specified and later grouped asacycline based or not. Because 81% of the chemo-y received was anthracycline based, we focusedthracycline-based regimens. Axillary surgeries wereined into three categories: SLNB only, SLNBed by ALND, or ALND only.
ge at diagnosis. Stage at diagnosis was classifieding to the American Joint Commission on Cancer
ncer Epidemiol Biomarkers Prev; 19(11) November 2010 2735



(AJCCable,stage

AnalyEac

(0 thrdiagntreatmverteddate.lymphof folnosednodegery,no loarm sFor

first overe.lymphsure fdate,lifestyextenwe comentinitiat

StatisInci

demahazarsures,PersostudyexposdevellymphWe

variabstandwithvariabthe saaddeddepensequeassociclinicatweenof asson suber thelymphremovwere

Wedescrion asequetreatmno raddemacumugenerysis,mode12-17,Pot

our mlympour uconfotreatmincludmodephedea serietribut

Resu

Welocatcountfromstudyrando(41%)the nocoopeconsecontaphysi(8%).to be icompwith sdate aentireaveraviewOf theup, 8557% cDem

potenat brewithquentpositilymp

Norman et al.

Cance2736

D


) criteria (35). If needed information was not avail-such as whether a metastatic workup was done,was coded as unknown.

sis planh woman's follow-up was divided into monthsough 59) from the reference date (date of histologicosis of breast cancer). Occurrences of subsequentent exposures or lymphedema events were con-from calendar time into months from the referenceWe defined end of follow-up as development ofedema, loss to follow-up, or completion of 5 yearslow-up. If, during follow-up, a woman was diag-with breast cancer or had a mastectomy or lymphsurgery on the side opposite to the original sur-she was censored at that time because there wasnger an unaffected side for comparing hand andizes.most analyses, the outcome was the time to theccurrence of any lymphedema, mild or more se-We also restricted the outcome to moderate/severeedema in sensitivity analyses. The effective expo-or all patient characteristics present at the referencesuch as medical conditions and demographic andle factors, was assumed to begin at month 0 andd throughout all months of follow-up. Similarly,nsidered the potential effects of each cancer treat-on lymphedema risk to be lasting, continuing fromion of that treatment to the end of follow-up.

tical analysisdence rate ratios [hazard ratios (HR)] of lymphe-(yes/no) were estimated using Cox proportionalds models to accommodate time-varying expo-censoring events, and loss to follow-up (36, 37).

n-months of a given exposure were combined oversubjects to estimate risk associated with that

ure. Only exposures occurring before lymphedemaoped were counted in evaluating risk factors foredema.used twomain analytic approaches: standard multi-le analyses with no interaction terms (referred to asard multivariable analyses throughout) and modelsinteraction terms. For example, standard multi-le analyses treat person-months of exposure to SLNBme regardless of whether other treatments have beenover time. However, the role of SLNB might differding on whether it is the sole exposure or a part of ance of exposures. Thus, tomodel risk of lymphedemaated with combinations of exposures, we includedlly plausible two- and three-way interactions be-treatments, asking, for example, whether the degreeociation between SLNB and lymphedema dependedsequent chemotherapy and/or radiation, or wheth-degree of association between radiation therapy andedema depended on having had prior lymph node
al and/or chemotherapy. Levels of these exposurescollapsed when necessary to achieve model fit.
race,educa

r Epidemiol Biomarkers Prev; 19(11) November 2010


used two different types of interaction models tobe the role of exposure sequence: HRs, which showratio scale the relative contribution of treatmentnces to risk of lymphedema compared with noent (no axillary surgery, no chemotherapy, andiation), and then cumulative incidence of lymphe-according to treatment. To estimate and plotlative incidence, we implemented an equivalentalized linear model for discrete time survival anal-but with a prespecified set of times to maintainl parsimony: each month 0-5 and months 6-11,18-23, and 24 or more, grouped as shown.ential confounding was addressed by including inodels factors previously reported as risk factors forhedema, those significant or close to significant innadjusted analyses, or those that seemed to act asunders in stratified analyses. Because protocols forent were based on stage at diagnosis, we did note stage and treatment in our models. Rather, weled the associations of cancer treatments and lym-ma risk, our primary objective, and then performeds of sensitivity analyses to clarify the separate con-ions of stage and treatment to lymphedema risk.

lts

ascertained 4,551 breast cancer cases from hospitals,ing 97% of breast cancer diagnoses in the twoies during the study period (38). The median timebreast cancer diagnosis to ascertainment for thewas 2 months (range, 0-33 months). Among 1,589mly selected potentially eligible patients, 649were enrolled. Patient refusal represented 25% ofnresponse; the remainder included physician non-ration (35%), inability to locate a physician to givent (8%), restrictive hospital requirements for patientct (13%), death (6%), illness (3%), ineligibility due tocal or mental incapability (3%), and inaccessibilityEighteen enrolled study subjects were discoveredneligible at the start because therewas no unaffectedarison side to evaluate lymphedema (17 patientsimultaneous bilateral mastectomies at the referencend 1 with a preexisting size difference affecting thearm), resulting in 631 patients in the study. Thege time from breast cancer diagnosis to first inter-was 12.2 months (median, 11; range, 1-28 months).631 study subjects, 94% completed 1 year of follow-% completed 2 years, 69% completed 3 years, andompleted all 5 years (29).ographic, lifestyle, and medical characteristics

tially related to lymphedema that were presentast cancer diagnosis are given in Table 1, alongunadjusted associations of each factor with subse-lymphedema. Statistically significant (P < 0.05)

ve associations with the presence and degree ofhedema were observed for younger age, black
higher body mass index (BMI), lower levels oftion and income, Medicaid/public assistance/no
Cancer Epidemiology, Biomarkers & Prevention



healthmedicMo

chemobreastThe pto lymat diademawe fipotenTablemateand lySev

with linsuralonge(Table(95%tiagen1.04-2ated wvariabfollow(HR,supra95% CobservsurgeTo f

indepmarknumbdiagnlymphen wiwerepropoalso twomethe sa5-year(39). Lwomreceivanthralympmoderseverestage Iunknopercenwas csevereWh

of theand l

stage,dencesampstageIIb/IIPar

cyclinthe nshowthe nu(HR,95% Cnodesare vible mnot inCI, 1.tion b(adjusHR

that atherapcompreferethat ismothsever4- toparedexcludALNDwithNonefromPlo

demathe fifindining tofour cregarelevatcurvewiseincidetwo cwiththeraptherapSLNBseemsin thesignif

Discu


www.a

D


insurance, greater perceived difficulty gettingal care, and stage at diagnosis.re than 98% of lymph node surgery, breast surgery,therapy, and radiation began in the first year aftercancer diagnosis, as did 93% of hormonal therapy.revalence of cancer treatments potentially relatedphedema is shown in Table 2 overall and by stagegnosis. When modeling the association of lymphe-with patient characteristics and cancer treatment,rst adjusted for potential confounding ignoringtial interactions (standard multivariable analyses;3) and then introduced interaction terms to esti-the association of specific treatment sequencesmphedema (Table 4; Fig. 1).eral patient characteristics significantly associatedymphedema without adjustment (age, race, healthnce, and perceived difficulty getting care) were nor significant after adjustment by all other factors3). ALND [HR, 2.61; 95% confidence interval

CI), 1.77-3.84] and chemotherapy, specifically mul-t therapies with anthracycline (HR, 1.46; 95% CI,.04), were the only treatments significantly associ-ith increased lymphedema risk in standard multi-le analyses. Risk was not significantly increaseding irradiation, whether to breast/chest wall only1.18; 95% CI, 0.80-1.73) or breast/chest wall plusclavicular nodal field +/- full axilla (HR, 0.86;I, 0.48-1.54). No significantly increased risk wased for hormonal therapy, SLNB, or type of breast

ry (Table 3).urther assess whether chemotherapy represented anendent risk factor for lymphedema, as opposed to aer for more aggressive disease, we performed aer of sensitivity analyses, first stratifying by stage atosis and then stratifying by number of positivenodes regardless of stage. The proportions of wom-

th stages IIa, IIb, and III/IV receiving chemotherapysimilar, around 80% (Table 2), and among these, thertion with anthracycline-based chemotherapy washe same, about 85% to 95%. The proportion ofn experiencing lymphedema was approximatelyme for stages IIa, IIb, and III/IV (Table 1), althoughrelative survival rates by stage differed markedlyymphedema did not occur as frequently among

en diagnosed at stage I (Table 1) but only 27%ed chemotherapy (Table 2). Among women takingcycline chemotherapy, the percentages experiencinghedema were, for stage I, 58% (37% mild, 21%ate/severe); stage IIa, 56% (31% mild, 24% moderate/); stage IIb, 58% (25% mild, 33% moderate/severe);II/IV, 47%(26%mild, 21%moderate/severe); and stagewn, 45% (26% mild, 19% moderate/severe). Thesets did not differ significantly whether lymphedemaategorized as any or none (P = 0.63) or as moderate/, mild, or none (P = 0.58).en we repeated our original multivariable analyses
association between anthracycline chemotherapy
ymphedema (Table 3), stratified, in addition, byOu

treatm

Caacrjournals.org


adjusted HRs remained elevated, although confi-intervals were wide in some strata due to limited

le size [stage I (n = 182): HR, 2.52; 95% CI, 1.33-4.76;IIa (n = 102): HR, 1.99; 95% CI, 0.82-4.81; stageI/IV (n = 73): HR 1.95; 95% CI, 0.55-6.89].allel adjusted analyses of the association of anthra-e chemotherapy and lymphedema controlling forumber of positive lymph nodes instead of stageed similar elevated HRs for chemotherapy, whethermber of positive nodes was treated as a continuous1.50; 95% CI, 1.06-2.11) or categorical (HR, 1.44;I, 1.01-2.06) variable, with the number of positivegrouped as 0, 1-3, 4-9, and 10 or more. These HRsrtually identical to the estimate from the compara-odel for chemotherapy and lymphedema that didclude number of positive nodes (HR, 1.46; 95%04-2.04; Table 3). Conversely, we found no associa-etween number of positive nodes and lymphedemated HR per positive node, 0.98; 95% CI, 0.93-1.03).s for treatment combinations from a complex modelllowed interactions of lymph node surgery, chemo-y, and radiation are shown in Table 4. All hazardsare the indicated combination of therapy against ance group of person-months with no risk factors,, no axillary surgery (no ALND or SLNB), no che-erapy, and no radiation (row I, Table 4). Overall,al treatment scenarios resulted in approximately5-fold increases in risk of lymphedema when com-with no treatment, with confidence intervals thated 1.0 (Table 4, rows A-D, F, and H). All involvedor chemotherapy or both, including combinations

SLNB and chemotherapy (Table 4, rows F and H).of the HRs for these categories differed significantlyeach other.ts of standardized cumulative incidence of lymphe-for six representative treatment combinations overrst 36 months of follow-up (Fig. 1) reinforce thesegs, suggesting two broad categories of risk accord-treatment combinations. As shown in the upperurves, cumulative incidence is highest for ALNDdless of subsequent treatment, but is similarlyed for SLNB followed by chemotherapy, and theses do not differ significantly, with P values for pair-comparisons ranging from 0.27 to 0.99. The lowestnce occurs after SLNB and no chemotherapy (lowerurves), with significantly lower incidence comparedthe highest group: P = 0.012 for SLNB, no chemo-y, radiation and P = 0.006 for SLNB, no chemo-y, no radiation. Within this low-risk category ofwithout chemotherapy, lymphedema incidenceto be higher when radiation therapy is includedtreatment regimen, but the results do not differ

icantly (HR, 5.77; 95% CI, 0.70-47.79).

ssion

r results highlight the relevance of breast cancerent types and patterns to lymphedema risk and




Tableperso

ALL elTota

Age (y<5050-780+

RaceWhiBlac

BMI<2525-230-335+

Educa<HigHighHighCollPos

MaritaMarDivoWidUnmNev

Type oCanTeaCom

Modifi01-23+

HypertYesNo

ArthritYes

ROU

No

SmokiNevEx-sCur

Norman et al.

Cancer E2738

Dow


me istics nt o
1. Demographic, lifestyle, and ns within levels of each factor; and ocia ema
81

22

35

80

1531

83

8104

84

94

52

81

3525

96

87

(Co

pidemiol Biomarkers Prev; 19(11) Novemb

ocebp.aacrjournals.org nloaded from

dical characterunadjusted ass

12.8 76.5

35.2 55.0

37.2 67.7

12.7 53.8

18.2 60.036.6 58.0

13.2 78.3

44.5 61.216.5 53.8

13.3 67.9

30.7 61.3

8.2 53.8

60.4 63.5

37.2 59.110.6 44.0

62.8 64.1

45.5 62.0

ntinued on the follow

er 2010

n June 1, 2018. © 20

potentially related to lymphedema; percetions of potential risk factors with lymphed

14.8

26.1

22.1 0.002

23.8

23.5 0.00529.4

15.7

28.1 0.1226.9

19.0

28.4

25.0

23.6

27.2 0.3136.0

23.2

23.3 0.45

ing page)

Cancer Epidemiology, Biomarkers & Preventi


f

Persons Lymphedema P*n % None% Mild% Moderate/severe%

igible womenl 631 100.0 62.3 24.7 13.0 —

)
1 18.4 < 96 31.1 54.6 27.0 0.001
9 354 56.1 63.3 25.7 11.0
8.6
te + others 409 64.8 66.3 24.0 9.8 0.001
k 2 18.9
2
10.2 9.99 2 12.0 00 31.7 64.0 24.04.99 116 18.4 54.3 31.9 13.8
22.5

tion level
h school 1 16.5 school graduate 2 12.6 school + some college 1 16.4 40 22.2 60.7 22.9
ege graduate 62 9.8 64.5 25.8 9.7
tgraduate 6.0
l status
ried 2 10.7 rced or separated 1 19.2 owed 1 11.6 46 23.1 67.8 20.5arried couple 16 2.5 56.3 18.8 25.0 er married 13.1
f hospital†

cer center 1
10.4 15 18.2 66.1 23.5 0.42ching 322 51.0 61.5 23.0 15.5 munity 1 10.3
ed Charlson index‡

3
13.7 51 55.6 64.1 22.2 0.27228 36.1 61.4 28.5 10.1
21.2

ension
250 39.6 60.4 26.4 13.2 0.55 3 12.9
is
2 13.6
heumatoid
20.0 steoarthritis only 7.8 90 38.3 63.3 28.9nknown type 120 51.1 59.2 24.2 16.7
3
12.6
ng status
er smoked 2 14.6 moker 194 30.7 63.9 26.8 9.3rent smoker 150 23.8 60.7 24.7 14.7
on


raisewithinviewfor lyparticWe fo

multitreatmphederecent

Tableperso and ocia ema(Cont'd)

DrinkinNon 36 53.2 62.8 23.5 0.45<1 d1 to7+ d 59 9.4 66.1 25.4

HealthPriv 38 53.6 62.4 24.0 0.005SomMedOth 38 6.0 60.5 21.1

DifficuExtrSomNot 70 74.5 64.7 24.0

House<$1 67 10.6 46.3 35.8 0.003$10 62 9.8 64.5 24.2$15 53 8.4 58.5 26.4$20 58 9.2 56.9 34.5$25 58 9.2 56.9 24.1$35 95 15.1 63.2 17.9$50$75Unk 60 9.5 71.7 20.0

StageIn s 84 13.8 82.1 14.3 0.001Stag 82 30.0 61.5 30.2Stag 02 16.8 47.1 32.3SSU

*TesMan†Am‡Movascmod§AJmordeswomby t hilade


www.a

D


questions for future research. Reviews publishedthe last 10 years have reinforced the prevailing

that the most important treatment risk factorsmphedema are ALND and radiation therapy,

he Pennsylvania Cancer Registry for the counties (Delaware and P

ularly radiation involving the axilla (7-15, 40).und that ALND and chemotherapy, specifically

of lymThe m

Caacrjournals.org


agent therapies with anthracycline, were the onlyents significantly associated with increased lym-ma risk in standard multivariable models. Fourly published studies also reported increased rates

lphia) and years (1999-2001) covered by our study (70).

1. Demographic, lifestyle, and medical characteristics potentially related to lymphedema; percent of
ns within levels of each factor; unadjusted ass tions of potential risk factors with lymphed
phedema associated with chemotherapy (41-4echanism by which chemotherapy would increa

ncer Epidemiol Biomarkers Prev; 19(11) November 20


Persons Lymphedema P*n % None% Mild% Moderate/severe%

g status
drinker 3 13.7 rink/wk 1 15.0 20 19.0 60.0 25.0<7 drinks/wk 116 18.4 61.2 27.6 11.2 rinks/wk 8.5 coverage ate insurance only 3 13.6 e Medicare 2 9.1 09 33.1 67.5 23.4icaid/public assistance/none 46 7.3 39.1 39.1 21.7 er/unknown 18.4 lty getting care emely/very 25.0 24 3.8 50.0 25.0 0.008ewhat/not very 135 21.4 56.3 26.7 17.0 at all 4 11.3 hold income 0,000 17.9 ,000-$14,999 11.3 ,000-$19,999 15.1 ,000-$24,999 8.6 ,000-$34,999 19.0 ,000-$49,999 18.9 ,000-$74,999 13.9 79 12.5 55.7 30.4,000+ 99 15.7 78.8 16.2 5.1 nown 8.3 at diagnosis§
itu
3.6 < e I 1 8.2 e IIa 1 20.6 e IIb 28.9 tag 45 7.4 48.9 22.2
tage III/IV 28 4.6 57.1 25.0 17.9nknown 166 27.3 63.3 21.7 15.1

ts of significance of association between lymphedema (none, mild, moderate/severe) and levels of exposures are based ontel-Haenszel χ2 tests for categorical or ordinal variables as appropriate.erican Association of Medical Colleges (68).dified Charlson index (69). Sum of scores, as assigned: Assign 1—myocardial infarction, congestive heart failure, peripheralular disease, stroke (no paralysis), chronic pulmonary disease, ulcer, mild liver disease, and diabetes; Assign 2—hemiplegia,erate or severe renal disease, diabetes with end organ damage, and any cancer other than breast cancer.CC stage (35). Complete AJCC staging information was not available for 166 patients because we lacked information on one ore components of stage. There were another 24 patients, who lacked information to distinguish between stage I or IIa. Except forcriptive analyses in Tables 1 and 2 that are stratified by stage, all study subjects were included in the analyses. Among the 441en with known AJCC stage in our study, the distribution of in situ (19%) and invasive (81%) cancers was the same as reported

4).se

10 2739


lymphchemdiseanodeslympthe caever,assoclymphat diacredibma aswhenmodegroup≥6 min thewometherapas taxpies [cizumevaluregard

Theshowsampas teapersisticed.the risubsechemhowetheratreatmof theanthrhapswithing aing SLlackedthesemid-1anthr

Table 2. Percent (number) of participants w

Overall g

i V

Any ra 4 1 7 3 1 8Amo%B 3 1 5 1 2 6%B 4 1 8 ( 8 1%B

fuAny ch 2 5 8 3 2 8

Amo% w

Any taAny ly 4 1 1 4 2 1

Amo%S%A%S

Any brAmo%M 1 3 3 1 1 7

%Any

*See†Th271‡Vir

Norman et al.

Cance2740

D


edema risk is not known, and it is possible thatotherapy might be a marker for more aggressivese. For example, greater tumor involvement inmight predispose to lymphedema by causing

hatic stasis for some period of time even beforencer is diagnosed and definitively treated. How-based on our sensitivity analyses, the positiveiation between anthracycline chemotherapy andedema does not seem to be confounded by stagegnosis or number of positive lymph nodes. Lendingility to our results, the increased rate of lymphede-sociated with anthracyclines was even strongerthe criterion for lymphedema was restricted torate/severe lymphedema (Table 3). In this sub-, 89% of women experienced lymphedema lastingonths. Anthracycline-containing regimens differedir constituent components, and the proportion ofn receiving a non–anthracycline-based chemo-y regimen in this cohort was small. Therapies, suchane without anthracycline, and other newer thera-e.g., trastuzumab (Herceptin), lapitinib, and beva-ab] did not occur in sufficient numbers to be

tually 100% of participants had breast surgery.

ated. Overall duration of therapy could be a factorless of specific agents.

Likassoc



lower risk of lymphedema after SLNB has beenn in many clinical studies; our population-basedle of patients seen at community hospitals, as wellching hospitals and cancer centers, documents thetence of this effect when SLNB is more widely prac-The HRs for treatment combinations suggest thatsk of lymphedema with SLNB depends on thequent use of anthracycline chemotherapy. Withoutotherapy, lymphedema risk was low after SLNB;ver, the risk of lymphedema after SLNB and chemo-py was significantly elevated compared with noent. As shown in Table 4, in our study, only 25%person-months of follow-up after SLNB involved

acycline chemotherapy, whereas 75% did not, per-explaining the overall lower risk of lymphedemaSLNB in this population. The many studies report-lower incidence of lymphedema in women receiv-NB compared with ALND (42, 45-54), for example,detailed information on chemotherapy. Most ofstudies recruited patients before or during the990s, and the proportions of patients receivingacycline would have varied considerably.

ith specifi
ed exposure, by stage of cancer at diagnosis
e ouiated

Ca

0 Am

Stage of ca

rs, other rewith radi

ncer Epide

erican Ass

ncer at dia

cent studieation thera

miology, B

ociation fo

nosis*

s found npy (41-4

iomarker

r Cancer R

o increase4). The re

s & Preve

esearch.

(n = 6
31) In s(n =
tu4)

I(n = 1
82)
IIa(n = 1
02)
IIb(n = 4
5)
III/I(n = 2
8)
DK/U(n = 1

nk*66)
8
diation 65.8 (
15) 53.6 (45) 78.0 ( 42) 68.6 ( 0) 77.8 ( 5) 64.3 ( 8) 53.0 ( 8) ng those with radiation: reast/chest wall only 80.2 (reast/chest wall + supraclavicular field 9.6 (
33) 100.0
(45) 99.3 ( 41) 77.1 ( 4) 42.9 ( 5) 11.1 ( ) 71.6 ( 3)
reast/chest wall + supraclavicular field + 10.1 (
0) 0.02) 0.0
(0) 0.7 ((0) 0.0 (

) 11.4 () 11.4 (

) 22.9) 34.3 (

8) 44.4 (2) 44.4 (

) 15.9 () 12.5 (

4)1)

ll axilla4 0 8 1 8 1

emotherapy 43.6 (
75) 1.2 (1)† 27.5 ( 0) 80.4 ( 2) 84.4 ( 8) 78.6 ( 2) 49.4 ( 2) ng those with chemotherapy: ith anthracycline 80.7 ( 22) 100.0 (1) 86.0 ( 3) 85.4 ( 0) 94.7 ( 6) 86.4 ( 9) 64.6 ( 3) 2 4 7 3 1 5moxifen 63.5 (401) 46.4 (39) 63.7 (116) 73.5 (75) 71.1 (32) 57.1 (16) 65.1 (108) mph node surgery 74.5 ( 70) 13.1 (11) 96.7 ( 76) 99.0 ( 01) 97.8 ( 4) 85.7 ( 4) 68.7 ( 14) ng those with any lymph node surgery: entinel + axillary 31.1 ( 46) 9.1 (1) 34.1 ( 0) 35.6 ( 6) 43.2 ( 9) 25.0 ( ) 21.1 ( 4) 1 6 3 1 6 2xillary only 50.2 (236) 63.6 (7) 36.4 (64) 53.5 (54) 54.5 (24) 66.7 (16) 62.3 (71)entinel only 16.6 (78) 18.2 (2) 29.5 (52) 10.9 (11) 0.0 (0) 4.2 (1) 10.5 (12) east surgery‡
ng those with any breast surgery:
astectomy 30.1 ( 90) 19.0 (16) 17.0 ( 1) 31.4 ( 2) 40.0 ( 8) 64.3 ( 8) 44.0 ( 3) with concurrent reconstruction 24.2 ( 6) 37.5 (6) 32.3 ( 0) 31.3 ( 0) 27.8 ( ) 22.2 ( ) 15.1 ( 1) % 4 1 1 5 4 1
Breast conservation 69.4 (438) 81.0 (68) 82.4 (150) 68.6 (70) 60.0 (27) 35.7 (10) 54.8 (91)breast/axilla infection 5.5 (35) 3.6 (3) 6.6 (12) 8.8 (9) 6.7 (3) 3.6 (1) 4.2 (7)

footnote § in Table 1.is participant first received chemotherapy in the third year after diagnosis. Of the 275 participants given chemotherapy,began chemotherapy within the first year of diagnosis.

d risklative

ntion


Tableand a

Poten

PatienAge gr

80+50-7

Race (Blac

BMI (r30+25-2

Educa<HigHigh

MaritaMar

HealthMedSom

DifficuExtrSom

TreatmLymph

SLNAny

SA

BreastMas

SN

ChemoAnthOth

RadiatBreaBreaBreafullBrea

(+Tamox

YBrea

Y

*The†Se‡Thefactors


www.aac

Dow


3. Patient and treatment risk factodjusted for all variables listed in the

tial risk factor*

to support estimation and interpretation.

rjournals.org

ocebp.aacrjournals.org nloaded from

rs for lymphedema. HRs at end oftable

Cancer Epidemiol B

n June 1, 2018. © 2010 American Asso

5 y of follow-up, unadjusted

ANY lymphedema, n =
238 Moder ate/severe lymphedem
iomarkers Prev; 19(11)

ciation for Cancer R

a, n = 82†

Unadjusted HRs(95% CI)

Adjusted HRs(95% CI)

Unadjusted HRs(95% CI)

Adjusted HRs(95% CI)

t characteristics
oup (reference: <50 y)
0.49 (0.30-0.80) 0.79 (0.43-1.46) 0.49 (0.22-1.11) 1.09 (0.38-3.14)
9 0.74 (0.56-0.97) 0.83 ( 0.60-1.14) 0.56 (0.35-0.88) 0.78 ( 0.45-1.36) reference: white) k 1.45 (1.12-1.88) 1.17 ( 0.87-1.57) 2.10 (1.36-3.24) 1.39 ( 0.86-2.25) eference: <25)
1.66 (1.23-2.26) 1.45 (1.04-2.02) 1.79 (1.06-3.02) 1.61 (0.91-2.83)
9.99 1.20 (0.87-1.66) 1.24 ( 0.89,1.74) 1.21 (0.69-2.13) 1.30 ( 0.71-2.37) tion level (reference: college graduate+) h school grad 1.72 (1.12-2.62) 1.49 (0.92-2.40) 2.61 (1.24-5.50) 1.84 (0.78-4.32) school grad or some college 1.69 (1.19-2.39) 1.56 ( 1.08-2.24) 2.02 (1.05-3.86) 1.79 ( 0.91-3.52)
l status (reference: unmarried)
ried 1.01 (0.79-1.31) 1.03 ( 0.78-1.36) 0.67 (0.43-1.05) 0.68 ( 0.42-1.11) coverage (reference: private insurance) icaid/public assistance/no insurance 2.22 (1.47-3.34) 1.46 (0.91-2.32) 1.92 (0.97-3.81) 0.93 (0.43-2.00) e Medicare 0.88 (0.65-1.18) 1.00 ( 0.69-1.44) 0.69 (0.41-1.18) 0.94 ( 0.47-1.85) lty getting care (reference: not at all) emely/very difficult 1.85 (1.03-3.33) 1.20 (0.61-2.35) 2.91 (1.25-6.79) 1.65 (0.61-4.51)ewhat difficult 1.44 (1.07-1.94) 1.25 (0.91-1.71) 1.73 (1.06-2.82) 1.52 (0.91-2.54)
ents
node surgery (reference: no lymph node su rgery) B only 1.08 (0.61-1.92) 1.04 ( 0.58-1.88) 0.99 (0.31-3.20) 0.99 ( 0.29-3.40) ALND 3.12 (2.20-4.43) 2.61 ( 1.77-3.84) 3.96 (1.99-7.86) 2.59 ( 1.17-5.74) LNB and ALND 2.89 (1.93-4.30) ‡ 3.18 (1.48-6.86) ‡
LND only 3.27
(2.27-4.73) ‡ 4.46 (2.21-9.02) ‡
surgery (reference: breast conservation)
tectomy 1.25 (0.94-1.65) 1.11 ( 0.78-1.58) 2.18 (1.40-3.41) 1.60 ( 0.90-2.86) imultaneous reconstruction 1.54 (0.96-2.45) ‡ 2.04 (0.96-4.34) ‡
o simultaneous reconstruction 1.16
(0.85-1.59) ‡ 2.23 (1.38-3.60) ‡
therapy (reference: none)
racycline-based regimen 2.11 (1.59-2.79) 1.46 (1.04-2.04) 5.23 (3.09-8.85) 3.76 (2.01-7.04) er/unknown type 1.49 (0.89-2.48) 1.05 ( 0.62-1.80) 4.37 (1.99-9.56) 3.16 ( 1.38-7.23) ion therapy (reference: none) st/chest wall only 1.02st/chest wall + supraclavicular field 1.32
(0.76-1.38) 1.18 (
0.80-1.73) 0.60 (0.35-1.04) 0.89 ( 0.43-1.81)
st/chest wall + supraclavicular field +axilla

1.33
(0.60-2.87) 0.77 ((0.66-2.68) 0.95 (
0.34-1.73) 2.130.46-1.97) 1.72

(0.92-4.93) 1.06 ((0.70-4.18) 1.03 (

0.44-2.60)0.40-2.69)

st/chest wall + supraclavicular field 1.33 (0.77-2.29) 0.86 (0.48-1.54) 1.92 (0.98-3.75) 1.05 (0.50-2.20)
/− full axilla). ifen (reference: none)
1.02
(0.76-1.36) 1.03 ( 0.76-1.39) 1.15 (0.71-1.87) 1.23 ( 0.73-2.08) esst/axillary infection (reference: none)es 1.80 (1.05-3.10) 1.15 (0.65-2.03) 1.11 (0.40-3.04) 0.70 (0.25-1.97)
numbers of patients in each category can be found in Tables 1 and 2.venty-three (89%) of the 82 women with moderate/severe lymphedema had lymphedema lasting 6 or more months.se sublevels were not included in the final model because there were too few observations to permit cross-classification of

November 2010 2741

esearch.


contriless evrecomally pservatboth cfirst. Otempomens,the stuAlt

treatmthe in(loweexposphedethe Hited acolleadiatiothat thTheirover wchangsequeSim

creasethe rewhich

potenanalythe asdid nRec

searchstudietion;demodetai(40). Cis obsthe dpreseof lymdomizstudynarromajorassessexposral orthe uslymphry meregimPot

Tableradia suradjusting for the potentially confounding patient and treatment factors

% of

(A) ALN(B) AL(C) AL(D) AL(E) SLN(F) SLN(G)(H) S(I) R

NOTreferesudevno c*Totbecanth†An‡Ra

Norman et al.

Cance2742

D


bution of radiation therapy to lymphedema may beident after chemotherapy. Consistent with currentmendations (24), in our study, chemotherapy gener-receded radiation for women undergoing breast con-ion. Ninety-one percent of the 210 patients receivinghemotherapy and radiation had their chemotherapyur results might be related to improvements in con-rary practice, not just changes in chemotherapy regi-but pursuing this further was beyond the scope ofdy.

hough incidence of lymphedema was lowest forents that did not include ALND or chemotherapy,teraction models presented in Table 4 and Fig. 1r two curves) suggested that in the absence of theseures, radiation might result in an increase in lym-ma risk. The curves seemed to be different andR for the comparison was large, but power was lim-nd results were not significant. Recently, Hayes andgues (55) reported that both chemotherapy and ra-n therapy increased the risk of lymphedema, ande risk increase was greater with axillary radiation.study recruited patients from 1970 to 2005, a periodhich approaches to radiation and chemotherapyed dramatically, and the effects of treatmentnce and type of chemotherapy were not explored.ilar to other reports (9, 11, 13, 15, 56), we found in-d risks of lymphedema in women with BMI ≥30 at

diation therapy: any radiation therapy regardless of extent.

ference date compared with those with lower BMI,remained statistically significant after adjusting for

bias, cbias,



tial confounding by treatment. However, sensitivityses showed no effect modification by BMI, that is,sociation of any treatment type and lymphedemaot differ by BMI (data not shown).ommendations for improving survivorship re-have stressed the need for large, population-baseds representing the diversity of the survivor popula-information not only on treatments but on socio-graphic factors and underlying comorbidities; andled timelines for exposures and adverse eventsonsistent with these recommendations, our studyervational and population based; it encompassesiverse population and wide range of treatmentsnt in the community that could influence the riskphedema, which would not be feasible with a ran-ed trial. Additional strengths are the prospectivedesign; the relatively long 5-year follow-up; the

w recruitment time frame of the study, precludingchanges in treatment patterns; comprehensivement of potential risk factors; the ability to assignures to each month of follow-up so that the tempo-der of exposures and outcome could be assessed;e of a previously validated questionnaire to assessedema from patient self-report; and supplementa-dical record reviews to obtain specifics of treatmentens.ential limitations relate to selection bias, information

4. HRs for incidence of lymphedema for combinations of axillary surgery, chemotherapy, and
tion therapy compared with a reference group of no axillary
onfoundingwe started

Cancer E

0 American

gery, radiation, or chemotherapy,specified in Table 3

(
Person-months of exposure
all person-months with that exposure)*

D, no chemotherapy, no radiation: HR, 3.78 (95% CI, 2.17-6.58)
2,265 (10.9) ND, anthracycline†, no radiation: HR, 5.46 (95% CI, 2.97-10.01) 1,470 (7.1) ND, no chemotherapy, radiation‡: HR, 4.67 (95% CI, 2.48-8.83) 2,704 (13.0) ND, anthracycline, radiation: HR, 4.61 (95% CI, 2.43-8.73) 3.632 (17.5) B, no chemotherapy, no radiation: HR, 0.30 (95% CI, 0.04-2.27) 718 (3.5) B, anthracycline, no radiation: HR, 4.06 (95% CI, 1.32-12.45) 222 (1.1)
NB, no chemotherapy, radiation: HR, 1.74 (95% CI, 0.70-4.37)
1,714 (8.3) SLLNB, anthracycline, radiation: HR, 4.09 (95% CI, 1.43-11.76) 589 (2.8)eference: No ALND or SLNB, no radiation, no chemotherapy: HR, 1.0 3,277 (15.8)
E: Reported HRs are estimates from a model with risk factors nested (a method for examining interaction terms). With arence group of subjects with no radiation, no chemotherapy, and no axillary surgery (no ALND or SLNB), each of the letteredlts (A-H) must be contrasted with this reference (row I). Thus, for example, the reported HR for (A) 3.78 is the relative hazard ofeloping lymphedema in those having ALND versus having no axillary surgery (no ALND or SLNB), assuming that the patient hadhemotherapy or radiation.al number of person-months in the model: 20,778. The number of person-months in the last column of the table totals to 16,591ause results for some treatment combinations with very few person-months of observation (e.g., chemotherapy but notracycline-based) are not shown in the table.thracycline chemotherapy: any regimen containing anthracycline.

, and sample size. To minimize selectionwith nearly complete enumeration of

pidemiology, Biomarkers & Prevention

Association for Cancer Research.


newlywell-dtentiaEnrolbut wfor yoof enrcer wfootnorestriphysihavetion wnoncoprivacPortabing dpotenthey weratiobe locFurthfoundestimaReg

was g81% bbetwe(29). TresultbegananalyReaso

loss thigheFin

modethat mTableimbalReg

our sttientcumfemethodemaaccepditionusedvalidativityhigh cbasedof ouruse ointer-It i

acrossing alengthouslyour stparabto 54%

FigureincidenaccordscenarisurvivarepresesubtracenhancplottedAnthraccontainRadiatitherapy


www.a

D


diagnosed breast cancer patients residing in ourefined study area and then randomly selected po-l participants. Among these, 41% were enrolled.lment was similar for white and black women,as lower for women ages 80 years and older thanunger women (data not shown). The proportionsolled women with in situ and invasive breast can-ere the same as in the population as a whole (seete in Table 1). Nonresponse resulted mainly fromctive hospital requirements for patient contact,cian refusal, or inability to find the physician. Weno evidence that physician/hospital noncoopera-as related to the risk of lymphedema. Instead,operation reflected global concerns about patienty, especially salient as the new Health Insuranceility and Accountability Act regulations were be-iscussed (29). Removing from the denominator oftially eligible patients those not enrolled becauseere inaccessible (i.e., physician/hospital noncoop-

n, died before initial contact, moved or could notated), the percent enrolled increased to 62% (29).er, some of the nonrespondents may have beento be ineligible had they enrolled, increasing theted response rate further.arding loss to follow-up, yearly retention overallenerally high, near or above 90%, dropping toy the third year of follow-up, the transition yearen the original study and continuation fundinghe longer time between contacts during this yeared in more losses to follow-up. Most lymphedemaduring the first 2 years of follow-up, and survival
sis maximized the use of follow-up months.ns for dropping out varied, with
measu

regardless of extent.

Caacrjournals.org


o follow-up occurred predominantly in persons atr or lower risk of lymphedema.ally, as described in Materials and Methods, allls were adjusted for patient and treatment factorsight influence the risk of lymphedema (listed in3), helping to reduce potential bias resulting fromances due to nonresponse or loss to follow-up.arding information bias, a potential limitation ofudy is that for practical reasons, we relied on pa-self-reports instead of directly measuring arm cir-rence or volume. However, the most appropriated to assess the presence and degree of lymphe-has been debated for years, and it is increasinglyted that the patient herself can best judge her con-(31-33, 57-59). In our study, trained interviewers

a structured questionnaire that we developed andted for this purpose. In the validation study, sensi-and specificity of self-reported lymphedema wereompared with expert physical therapists' diagnosison arm measurements (28, 29). Another advantageapproach is that, unlike direct arm measurements,

f a structured questionnaire minimizes intra- andobserver variability (60).s difficult to compare incidence of lymphedemastudies, given the different approaches to measur-

nd defining lymphedema, as well as the variables of follow-up. Nonetheless, as we reported previ-(29), the 35% incidence of lymphedema at 3 years inudy was in the range reported by others for a com-le time period, varying from 15% (61) to 21% (62, 63)(42). The first two studies used circumferential

rements but with different criteria for lymphedema
no indication that (61, 62), and the third and fourth used self-reports (42, 63).
1. Estimated cumulativece of lymphedemaing to different treatmentos based on discrete timel models. The jitternts a small addition/tion to the true values toe separation of thevalues for visual clarity.ycline therapy: any regimening anthracyclines.on therapy: any radiation




Bypercesomelympcombheaviment51, 53devellymphity oflymphwereeffectsseparaPre

treatmcaptuand lreporof bremedicchemfieldsPot

widefactoras coconfofoundwerenot awereof suweighturn tincreanism.were u

Anstudyanalysupraquesttheraprisk. Fdemawith rresultsufficfinerAlt

a singof othclarifysel andemaconsising ddifferframebehind the association between cancer treatments andlymph

Disclosure of Potential Conflicts of Interest

No p

Grant

NatiSociety

Thepaymenadvertisthis fac

Refe1. Pa

lym83:

2. Tomomo

3. Mcary

4. Kuaxisiglog275

5. Paim

6. Fulym38

Norman et al.

Cance2744

D


design, our measure of lymphedema relies solely onived differences in size between the limbs, unlikeother questionnaires developed for self-reportedhedema assessment in which swelling has beenined with additional factors, such as discomfort,ness, tightness, decreased functional activity, move-limitation, and need for a compression sleeve (50,, 63). An important objective of our questionnaireopment was to design a questionnaire that definededema independent of its potential effect on qual-life. This was in keeping with the definitions ofedema at the time we designed our study, whichbased on size. Further, to learn more about theof lymphedema on quality of life, we needed tote size from interference with daily activities (29).sence and timing of medical conditions and cancerent were also obtained from self-reports to bestre the relative time difference between exposuresymphedema outcome. Agreement between self-ts and medical records for the presence and datesast cancer treatment is high (64-67), and we usedal records for treatment details such as types ofotherapy, lymph node dissection, and radiationthat have been shown to be less well remembered.ential confounding was addressed by examining arange of exposures previously reported as risks for lymphedema as well as those that might actnfounders within our data. When evaluatingunding, we distinguished between potential con-ers (independent risk factors for lymphedema thatassociated with the exposure of interest but wereconsequence of the exposure) and other factors thatintervening variables, not confounders. An examplech a potential physiologic mechanism might bet gain subsequent to chemotherapy, which inriggers lymphedema. Chemotherapy would stillse the risk of lymphedema regardless of the mecha-
Finally, as with all studies, it is possible that there
hatwere unaccounted for.Rece

publish

MR, Rosedale M. Breast cancer survivors' experiences ofphedema-related symptoms. J Pain Symptom Manage 2009;:849–59.

7. Arman

8. Eriin

9. Koca

10. Pabre

11. MoBre

12. Sptre

13. Wisiz

14. Peres

15. Rorea



other limitation relates to sample size. Although ourwas large, we lacked power in some subgroup

ses. The number of women with radiation to theclavicular fields was small, leaving some openions about the association of the extent of radiationy and lymph node irradiation with lymphedemaor example, it seemed, anomalously, that lymphe-risk was lower with more extensive radiation thanadiation limited to the breast/chest wall, but theses did not differ significantly. The study also lackedient power to study the types of chemotherapy indetail.hough our study suggests that the increase in risk ofle therapy may depend on the presence or absenceer therapies, more research is needed to replicate,, and extend these findings to appropriately coun-d monitor breast cancer survivors about lymphe-risk associated with breast cancer treatment. Thetency of recent studies, each with somewhat vary-efinitions of lymphedema and each conducted inent populations but over a relatively similar time, supports further pursuit of potential mechanisms

edema.

otential conflicts of interest were disclosed.

Support

onal Cancer Institute grant R01 CA65422 and American Cancergrant ROG-02-259 (S.A. Norman).costs of publication of this article were defrayed in part by thet of page charges. This article must therefore be hereby markedement in accordance with 18 U.S.C. Section 1734 solely to indicatet.

ived 12/10/2009; revised 07/06/2010; accepted 07/12/2010;ed OnlineFirst 10/26/2010.
nidentified confounders t
rencesssik SD, McDonald MV. Psychosocial aspects of upper extremityphedema in women treated for breast carcinoma. Cancer 1998;2817–20.bin MB, Lacey HJ, Meyer L, Mortimer PS. The psychologicalrbidity of breast cancer-related arm swelling: Psychologicalrbidity of lymphoedema. Cancer 1993;72:3248–52.Wayne J, Heiney SP. Psychologic and social sequelae of second-lymphedema: a review. Cancer 2005;104:457–66.ehn T, Klauss W, Darsow M, et al. Long-term morbidity followingllary dissection in breast cancer patients—clinical assessment,nificance for life quality and the impact of demographic, onco-ic and therapeutic factors. Breast Cancer Res Treat 2000;64:–86.skett ED, Stark N. Lymphedema: Knowledge, treatment andpact among breast cancer survivors. Breast J 2000;6:373–8.

er J. The problem of post-breast cancer lymphedema: impactd measurement issues. Cancer Invest 2005;23:76–83.ckson VS, Pearson ML, Ganz PA, Adams J, Kahn KL. Arm edemabreast cancer patients. J Natl Cancer Inst 2001;93:96–111.cak Z, Overgaard J. Risk factors of arm lymphedema in breastncer patients. Acta Oncol 2000;39:389–92.in SJ, Purushotham AD. Lymphoedema following surgery forast cancer. Br J Surg 2000;87:1128–41.rrell RM, HalyardMY, Schild SE, Ali MS, Gunderson LL, Pockaj BA.ast cancer-related lymphedema. Mayo Clin Proc 2005;80:1480–4.araco A, Fentiman IS. Arm lymphoedema following breast canceratment. Int J Clin Pract 2002;56:107–10.lliams AF, Franks PJ, Moffatt CJ. Lymphoedema: estimating thee of the problem. Palliat Med 2005;19:300–13.trek JA, Pressman PI, Smith RA. Lymphedema: current issues in
earch and management. CA Cancer J Clin 2000;50:292–307.ckson SG. Precipitating factors in lymphedema: myths andlities. Cancer 1998;83:2814–6.



16. MaBetio

17. Erbsen

18. Tjacomthe(pN(SN(su

19. Bilneldis294

20. GiudoN0

20121. Ev

finevi29

22. WhOsLip

23. BuwhOsLip

24. Pie199pama200

25. WhRarad5:7

26. BuDeanded

27. HaCoea

28. Nooplym119

29. Nocansym

30. WrdiaresCa

31. ArmcanRe

32. Ge280

33. Pelym

34. HuOn

35. Gre6th

36. Kle(CA

37. Kada

38. PeCaan

39. Amfrois_

40. BhanBio

41. GeciaHe

42. Paepbre16

43. MelymWh

44. Shcoof20

45. VesenN E

46. Pusea r

47. LyClibio

48. Matriaop20

49. FlemotriainCa

50. Blamoor48

51. Hafolca

52. Scmobio

53. Scsefor

54. Fraingsevaxi

55. Haboalo72

56. NawwHe


www.a

D


ggard MA, Lane KE, O'Connell JB, Nanyakkara DD, Ko CY.yond the clinical trials: how often is sentinel lymph node dissec-n performed for breast cancer? Ann Surg Oncol 2005;12:41–7.KM, Julian TB. Completion of axillary dissection for a positivetinel node: necessary or not? Curr Oncol Rep 2009;11:15–20.n-Heijnen VC, Pepels MJ, de Boer M, et al. Impact of omission ofpletion axillary lymph node dissection (cALND) or axillary radio-rapy (ax RT) in breast cancer patients with micrometastases1mi) or isolated tumor cells (pN0[i+] in the sentinel lymph node): results from the MIRROR study. J Clin Oncol 2009;27:18s,ppl;abstr CRA506).imoria KY, Bentrem DJ, Hansen NM, et al. Comparison of senti-lymph node biopsy alone and completion axillary lymph nodesection for node-positive breast cancer. J Clin Oncol 2009;27:6–53.liano AE, McCall LM, Beitsch PD, et al. ACOSOG Z0011: A ran-mized trial of axillary node dissection in women with clinical T1-2M0 breast cancer who have a positive sentinel node. J Clin Oncol0;28, (18s (suppl;abstr CRA506)).ans SB, Gass J, Wazer DE. Management of the axilla after theding of a positive sentinel lymph node: a proposal for andence-based risk-adapted algorithm. Am J Clin Oncol 2008;31:3–9.ite JR. Axillary irradiation. In: Harris JR, Lippman ME, Morrow M,borne CK, editors. Diseases of the breast. 4th ed. Philadelphia:pincott Williams & Wilkins; 2010, p. 570–7.chholz TA, Hunt KK. Breast-conserving therapy: conventionalole breast irradiation. In: Harris JR, Lippman ME, Morrow M,borne CK, editors. Diseases of the breast. 4th ed. Philadelphia:pincott Williams & Wilkins; 2010, p. 507–21.rce LJ, Moughan J, White J, Winchester DP, Owen J, Wilson JF.8-1999 patterns of care study process survey of national practice

tterns using breast-conserving surgery and radiotherapy in thenagement of stage I-II breast cancer. Int J Radiat Oncol Biol Phys5;62:183–92.ite JR, Halberg FE, Rabinovitch R, et al. American College ofdiology appropriateness criteria on conservative surgery andiation: stages I and II breast carcinoma. J Am Coll Radiol 2008;01–13.rstein HJ, Harris JR, Morrow M. Malignant tumors of the breast. In:Vita VT, Lawrence TS, Rosenberg SA, editors. DeVita, Hellman,Rosenberg's cancer: principles and practice of oncology. 8th

. Philadelphia: Lippincott Williams and Wilkins; 2008, p. 1606–54.rlan LC, Clegg LX, Abrams J, Stevens JL, Ballard-Barbash R.mmunity-based use of chemotherapy and hormonal therapy forrly-stage breast cancer: 1987-2000. J Clin Oncol 2006;24:872–7.rman SA, Miller LT, Erikson HB, Norman MF, McCorkle R. Devel-ment and validation of a telephone questionnaire to characterizephedema in women treated for breast cancer. Phys Ther 2001;81:2–205.rman SA, Localio AR, Potashnik SL, et al. Lymphedema in breastcer survivors: incidence, degree, time course, treatment, andptoms. J Clin Oncol 2009;27:390–7.ight B. Personal communication: Number of breast cancersgnosed in 1997 restricted to Philadelphia and Delaware Countyidents, by reporting hospital. Harrisburg (PA): Pennsylvaniancer Registry; 2000.er JM, Radina ME, Porock D, Culbertson SD. Predicting breastcer-related lymphedema using self-reported symptoms. Nurss 2003;52:370–9.rber LH. A review of measures of lymphedema. Cancer 1998;83:3–4.trek JA, Heelan MC. Incidence of breast carcinoma-relatedphedema. Cancer 1998;83:2776–81.ll M. Lymphedema in women treated for breast cancer. Semincol Nurs 2000;16:226–37.ene FL, Page DL, Fleming ID, et al. AJCC cancer staging manual.
ed. New York (NY): Springer-Verlag; 2002.inbaum DG, Kupper LL, H M. Epidemiologic research. Belmont): Lifetime Learning Publications; 1982.
57. Pewolym

Caacrjournals.org


lbfleisch JD, Prentice RL. The statistical analysis of failure timeta. New York: John Wiley and Sons; 1980.nnsylvania Department of Health. Cancer incidence 1990-2005.ncer cases by 23 sites for Pennsylvania and by county by sex, raced age. Harrisburg (PA): Pennsylvania Department of Health; 2008.erican Cancer Society. How is breast cancer staged? Availablem: http://www.cancer.org/docroot/CRI/content/CRI_2_4_3X_How_breast_cancer_staged_5.asp; 2009.atia S, Robison LL. Cancer survivorship research: opportunitiesd future needs for expanding the research base. Cancer Epidemiolmarkers Prev 2008;17:1551–7.ller BM, Vacek PM, O'Brien P, Secker-Walker RH. Factors asso-ted with arm swelling after breast cancer surgery. J Womensalth 2003;12:921–30.skett ED, Naughton MJ, McCoy TP, Case LD, Abbott JM. Theidemiology of arm and hand swell ing in premenopausalast cancer survivors. Cancer Epidemiol Biomarkers Prev 2007;:775–82.eske KA, Sullivan-Halley J, Smith AW, et al. Risk factors for armphedema following breast cancer diagnosis in Black women andite women. Breast Cancer Res Treat 2009;113:383–91.ih YC, Xu Y, Cormier JN, et al. Incidence, treatment costs, andmplications of lymphedema after breast cancer among womenworking age: a 2-year follow-up study. J Clin Oncol 2009;27:07–14.ronesi U, Paganelli G, Viale G, et al. A randomized comparison oftinel-node biopsy with routine axillary dissection in breast cancer.ngl J Med 2003;349:546–53.rushotham AD, Upponi S, Klevesath MB, et al. Morbidity afterntinel lymph node biopsy in primary breast cancer: results fromandomized controlled trial. J Clin Oncol 2005;23:4312–21.man GH, Giuliano AE, Somerfield MR, et al. American Society ofnical Oncology guideline recommendations for sentinel lymph nodepsy in early-stage breast cancer. J Clin Oncol 2005;23:7703–20.nsel RE, Fallowfield L, Kissin M, et al. Randomized multicenterl of sentinel node biopsy versus standard axillary treatment inerable breast cancer: the ALMANAC Trial. J Natl Cancer Inst06;98:599–609.issig A, Fallowfield LJ, Langridge CI, et al. Post-operative armrbidity and quality of life. Results of the ALMANAC randomisedl comparing sentinel node biopsy with standard axillary treatmentthe management of patients with early breast cancer. Breastncer Res Treat 2006;95:279–93.nchard DK, Donohue JH, Reynolds C, Grant CS. Relapse andrbidity in patients undergoing sentinel lymph node biopsy alonewith axillary dissection for breast cancer. Arch Surg 2003;138:2–8.id A, Kuehn T, Konstantiniuk P, et al. Shoulder-arm morbiditylowing axillary dissection and sentinel node only biopsy for breastncer. Eur J Surg Oncol 2002;28:705–10.hijven MP, Vingerhoets AJ, Rutten HJ, et al. Comparison ofrbidity between axillary lymph node dissection and sentinel nodepsy. Eur J Surg Oncol 2003;29:341–50.hrenk P, Rieger R, Shamiyeh A, Wayand W. Morbidity followingntinel lymph node biopsy versus axillary lymph node dissectionpatients with breast carcinoma. Cancer 2000;88:608–14.ncis WP, Abghari P, Du W, Rymal C, Suna M, Kosir MA. Improv-surgical outcomes: standardizing the reporting of incidence anderity of acute lymphedema after sentinel lymph node biopsy andllary lymph node dissection. Am J Surg 2006;192:636–9.yes SB, Freedman GM, Li T, Anderson PR, Ross E. Does axillaryost increase lymphedema compared with supraclavicular radiationne after breast conservation? Int J Radiat Oncol Biol Phys 2008;:1449–55.tional Cancer Institute. Lymphedema (PDQ®). Available from: http://w.cancer.gov/cancertopics/pdq/supportivecare/lymphedema/althProfessional/page2; 2008.
trek J, Pressman P, Smith R. Lymphedema: Results from arkshop on breast cancer treatment-related lymphedema and-phedema resource guide. Cancer 1998;83:2775–890.



58. RoLymof

59. Paica566

60. Armlym20

61. HeG.bre

62. ClafolQ

63. Ensevbre

64. Masuter

65. Phreptio46

66. Scbreep20

67. Liucabre

68. Aming

69. Chclame

70. PeRe

Norman et al.

Cance2746

D


ckson SG, Miller LT, Senie R, et al, American Cancer Societyphedema Workshop. Workgroup III: diagnosis and management

lymphedema. Cancer 1998;83:2882–5.skett ED. Breast cancer-related lymphedema: attention to a signif-nt problem resulting from cancer diagnosis. J Clin Oncol 2008;26:6–7.er J, Stewart B. A comparison of four diagnostic criteria forphedema in a post-breast cancer population. Lymphat Res Biol05;3:208–17.rd-Smith A, Russo A, Muraca MG, Rosselli Del Turco M, CardonaPrognostic factors for lymphedema after primary treatment forast carcinoma. Cancer 2001;92:1783–7.rk B, Sitzia J, Harlow W. Incidence and risk of arm oedemalowing treatment for breast cancer: a three-year follow-up study.J Med 2005;98:343–8.gel J, Kerr J, Schlesinger-Raab A, Sauer H, Holzel D. Axilla surgeryerely affects quality of life: results of a 5-year prospective study inast cancer patients. Breast Cancer Res Treat 2003;79:47–57.
unsell E, Drolet M, Ouhoummane N, Robert J. Breast cancerrvivors accurately reported key treatment and prognostic charac-istics. J Clin Epidemiol 2005;58:364–9.
Phapas



illips KA, Milne RL, Buys S, et al. Agreement between self-orted breast cancer treatment and medical records in a popula-n-based Breast Cancer Family Registry. J Clin Oncol 2005;23:79–86.hootman M, Jeffe DB, West MM, Aft R. Self-report by elderlyast cancer patients was an acceptable alternative to surveillance,idemiology, and end results (SEER) abstract data. J Clin Epidemiol05;58:1316–9.Y, Diamant AL, Thind A, Maly RC. Validity of self-reports of breast

ncer treatment in low-income, medically underserved women withast cancer. Breast Cancer Res Treat 2010;119:745–51.erican Association of Medical Colleges. Member institution list-s. Available from: http://services.aamc.org/memberlistings; 2010.arlson ME, Pompei P, Ales KL, MacKenzie CR. A new method ofssifying prognostic comorbidity in longitudinal studies: develop-nt and validation. J Chron Dis 1987;40:373–83.nnsylvania Department of Health. Pennsylvania Cancergistry Dataset. Percent staging distribution of breast cancer foriladelphia and Delaware Counties, 1999-2001. Available from: http://p2.health.state.pa.us/epiqms/Asp/SelectParams_Tbl_Cancer.
p; 2010.



2010;19:2734-2746. Published OnlineFirst October 26, 2010.Cancer Epidemiol Biomarkers Prev Sandra A. Norman, A. Russell Localio, Michael J. Kallan, et al. Risk Factors for Lymphedema after Breast Cancer Treatment

Updated version

10.1158/1055-9965.EPI-09-1245doi:

Access the most recent version of this article at:

Cited articles

http://cebp.aacrjournals.org/content/19/11/2734.full#ref-list-1

This article cites 57 articles, 11 of which you can access for free at:

Citing articles

http://cebp.aacrjournals.org/content/19/11/2734.full#related-urls

This article has been cited by 4 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cebp.aacrjournals.org/content/19/11/2734To request permission to re-use all or part of this article, use this link

on June 1, 2018. © 2010 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from


http://cebp.aacrjournals.org/lookup/doi/10.1158/1055-9965.EPI-09-1245

http://cebp.aacrjournals.org/content/19/11/2734.full#ref-list-1

http://cebp.aacrjournals.org/content/19/11/2734.full#related-urls

http://cebp.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cebp.aacrjournals.org/content/19/11/2734


& prevention risk factors for lymphedema after...

Documents