Full-Field Digital versus Screen- Film Mammography - … · Full-Field Digital versus Screen-Film...
Transcript of Full-Field Digital versus Screen- Film Mammography - … · Full-Field Digital versus Screen-Film...
Full-Field Digital versus Screen-Film Mammography: Comparisonwithin the UK Breast ScreeningProgram and Systematic Review ofPublished Data1
Sarah Vinnicombe, MRCP, FRCRSnehal M. Pinto Pereira, MScValerie A. McCormack, MSc, PhDSusan Shiel, BScNick Perry, MDIsabel M. dos Santos Silva, MD, PhD
Purpose: To (a) compare the performance of full-field digital mam-mography (FFDM), using hard-copy image reading, withthat of screen-film mammography (SFM) within a UKscreening program (screening once every 3 years) forwomen aged 50 years or older and (b) conduct a meta-analysis of published findings along with the UK data.
Materials andMethods:
The study complied with the UK National Health ServiceCentral Office for Research Ethics Committee guidelines;informed patient consent was not required, since analysiswas carried out retrospectively after data anonymization.Between January 2006 and June 2007, a London popula-tion-based screening center performed 8478 FFDM and31 720 SFM screening examinations, with modality deter-mined by the type of machine available at the screeningsite. Logistic regression was used to assess whether breastcancer detection rates and recall rates differed betweenscreening modalities. For the meta-analysis, random-ef-fects models were used to combine study-specific esti-mates, if appropriate.
Results: A total of 263 breast cancers were detected. After adjust-ment for age, ethnicity, area of residence, and type ofreferral, there was no evidence of differences betweenFFDM and SFM in terms of detection rates (0.68 [95%confidence interval {CI}: 0.47, 0.89] vs 0.72 [95% CI:0.58, 0.85], respectively, per 100 screening mammo-grams; P � .74), recall rates (3.2% [95% CI: 2.8, 3.6] vs3.4% [95% CI: 3.1, 3.6]; P � .44), positive predictivevalue (PPV) of an abnormal mammogram, or characteris-tics of detected tumors. Meta-analysis of data from eightstudies showed a slightly higher detection rate for FFDM,particularly at 60 years of age or younger (pooled FFDM-SFM difference: 0.11 [95% CI: 0.04, 0.18] per 100 screen-ing mammograms), but no clear modality differences inrecall rates or PPVs.
Conclusion: Within a routine screening program, FFDM with hard-copy image reading performed as well as SFM in terms ofprocess indicators; the meta-analysis was consistent withFFDM yielding detection rates at least as high as those forSFM.
� RSNA, 2009
1 From the Breast Unit, St Bartholomew’s Hospital, Bartsand The London NHS Trust, West Smithfield, LondonEC1A 7BE, England (S.V., S.S., N.P.); and Cancer Re-search UK Epidemiology and Genetics Group, LondonSchool of Hygiene and Tropical Medicine, London, En-gland (S.M.P.P., V.A.M., I.M.d.S.S.). From the 2007 RSNAAnnual Meeting. Received July 15, 2008; revision re-quested August 19; revision received October 14; ac-cepted November 5; final version accepted December 11.Address correspondence to S.V. (e-mail: [email protected] ).
� RSNA, 2009
ORIGINALRESEARCH
�BREAST
IMAGING
Radiology: Volume 251: Number 2—May 2009 ▪ radiology.rsnajnls.org 347
Note: This copy is for your personal, non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, use the Radiology Reprints form at the end of this article.
Randomized controlled trials haveshown that screening mammogra-phy reduces breast cancer mortal-
ity, largely through its ability to depictsubtle soft-tissue masses and microcalci-fications that may represent early breastcancer (1,2). These trials were based onthe use of screen-film mammography(SFM). Mortality has not been assessedfor other screening modalities; rather, re-liance has been placed on surrogate endpoints such as breast cancer detectionrates, recall rates, and tumor character-istics (eg, size and stage) at diagnosis.
Full-field digital mammography(FFDM) was approved by the U.S. Foodand Drug Administration in 2000, but itsincorporation into screening mammogra-phy programs has been slow. Fewer than10% of mammography units in the UnitedStates were digital in 2006 (3) and in theUnited Kingdom, where most mammo-graphic screening takes place within theNational Health Service Breast ScreeningProgramme (NHSBSP), the move toFFDM has been even slower. Major ob-stacles to FFDM include the high cost ofthe units and of archiving facilities and thelogistics of converting from analogue todigital systems. In the United Kingdom,most screening takes place in mobileunits that move around catchment areas,with film being developed and imagesread at the base hospital.
Authors of previous studies have at-tempted to compare FFDM to SFM, butfew have done so within an operationalnational screening program. The newpurpose-built Breast Unit at St Bar-tholomew Hospital (London, England),which houses the Central and East Lon-don Breast Screening Service (CELBSS),was the first NHSBSP unit in the UnitedKingdom to be equipped with FFDMunits in 2005, although thus far it has notbeen possible to implement soft-copy im-age reading in the screening setting.
The aim of our study was to comparethe performance of FFDM, using hard-copy image reading, with that of conven-tional SFM within the context of a UKmammography screening program forwomen aged 50 years or older (two-viewmammograms obtained once every 3years). In addition, a systematic literaturereview on the performance of FFDM rel-ative to that of SMF, along with the newUK data, was conducted to summarizethe available evidence.
Materials and Methods
CELBSS StudyThe CELBSS is covered by the generalethical approval of the NHSBSP. Ourstudy complied with the UK NationalHealth Service Central Office for Re-search Ethics Committee guidelines.Informed patient consent was not con-sidered necessary as this retrospec-tive analysis was carried out after dataanonymization.
The NHSBSP is a population-basedscreening program that began in 1988 andoffers two-view mammography once every3 years to women aged 50–70 years.Women older than 70 years can self-referevery 3 years. The CELBSS, one of the re-gional NHSBSP centers, has an eligible pop-ulation of 120 000 women. The program is
quality assured according to NHSBSPguidelines (4), with an average standard-ized detection ratio of 1.25 in 2006–2007.Uptake is, however, relatively low acrossthis deprived urban inner-city area (uptakeis 50% compared with the national mini-mum target of 70%).
A total of 40 198 screening examina-tions in 39 651 women were performedby the CELBSS between January 1, 2005,when FFDM was first implemented, andJune 30, 2007. The majority of screeningexaminations were performed in two mo-bile units, each equipped with an M-IVmammography unit (Lorad, London, En-gland). The remainder of the screeningexaminations were conducted in a staticunit at St Bartholomew’s Hospital, whichhas four FFDM units: three SenographeDS units (GE Healthcare, Slough, En-gland) and one Selenia unit (Lorad). Thescreening location, and thus modality,was mainly influenced by residential area,since women are invited for screening ac-cording to the location of their generalpractitioner’s practice. A small percent-age of women (eg, those with specialneeds or implants) are preferentially in-vited for screening at St Bartholomew’sBreast Unit, and those working nearbycan request to be screened in this unit.
Published online10.1148/radiol.2512081235
Radiology 2009; 251:347–358
Abbreviations:BI-RADS � Breast Imaging Reporting and Data SystemCELBSS � Central and East London Breast Screening
ServiceCI � confidence intervalFFDM � full-field digital mammographyNHSBSP � National Health Service Breast Screening
ProgrammePPV � positive predictive valueSFM � screen-film mammography
Author contributions:Guarantors of integrity of entire study, S.V., I.M.d.S.S.;study concepts/study design or data acquisition or dataanalysis/interpretation, all authors; manuscript drafting ormanuscript revision for important intellectual content, allauthors; approval of final version of submitted manu-script, all authors; literature research, S.V., S.M.P.P.,I.M.d.S.S.; clinical studies, S.V., N.P.; statistical analysis,S.M.P.P., V.A.M., I.M.d.S.S.; and manuscript editing, S.V.,S.M.P.P., V.A.M., I.M.d.S.S.
Authors stated no financial relationship to disclose.
Advances in Knowledge
� In a UK routine population-basedscreening program for womenaged 50 years or older (screeningevery 3 years), full-field digitalmammography (FFDM) usinghard-copy reading performed aswell as screen-film mammography(SFM) in terms of breast cancerdetection rates, recall rates, andpositive predictive value (PPV) ofan abnormal mammogram.
� A meta-analysis of data fromseven published studies, togetherwith those from our UK study,was consistent with FFDM havingdetection rates at least as high asthose for SFM, with no clear mo-dality differences in recall rates orPPVs.
Implication for Patient Care
� Digital mammography with hard-copy reading performs, in termsof process measures, at least aswell as conventional screen-filmmammography for breast cancerscreening.
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
348 radiology.rsnajnls.org ▪ Radiology: Volume 251: Number 2—May 2009
All mammography units have compa-rable focal spot size and grid and are sub-jected to rigorous quality control proce-dures as specified in the NHSBSP (4) andEuropean quality assurance guidelines(5). Images obtained at the mobile unitswere processed at the base unit. Readingof the SFM film images and hard-copyreading of the FFDM images took placewith the same multiloaders and with thesame viewing conditions. For the first 18months, the optimized postprocessed im-ages (“Premium View” from SenographeDS units; GE Healthcare) could not beprinted, but a software upgrade renderedthis possible starting in June 2006. Thus,the unoptimized images were read for thefirst 18 months. Images from the FFDMscreening examinations were printedwith an 8900 laser hard-copy unit(Kodak, Hemel Hempstead, England)(38.75-�m laser spot size, 650 dpi reso-lution).
Each SFM and FFDM image was readby two of six radiologists (a pool of six pos-sible readers: S.V., N.P., and four nonau-thors), all but one of whom had more than10 years of experience in breast screening,reading a minimum of 5000 screening
Table 1
Characteristics of Study Subjects by Screening Modality
CharacteristicNo. of SFMExaminations*
No. of FFDMExaminations* Difference†
Total no. of screening examinations 31 720 (100.0)‡ 8478 (100.0)‡
Type of screening or referral 10.1 (9.3, 10.9)§
Routine recall 21 547 (67.9) 4305 (50.8)Prevalent round 8352 (26.3) 2831 (33.4)General practitioner referral 147 (0.46) 37 (0.4)Self-referral 1674 (5.28) 1305 (15.4)
Age (y) 2.3 (1.2, 3.5)�60 18 939 (59.7) 5263 (62.1)�60 12 781 (40.3) 3215 (37.9)
Ethnicity 4.5 (5.7, 3.3)Caucasian 17 828 (64.6) 4945 (69.1)Non-Caucasian 9770 (35.4) 2212 (30.9)Missing data 4122 (13.0) 1321 (15.6)
No. of recalls 1404 (4.43) 406 (4.79) 0.36 (�0.15, 0.87)No. of detected breast cancers 205 (0.65) 58 (0.68) 0.04 (�0.16, 0.23)
* Data in parentheses are percentages.† Data in parentheses are 95% CIs.‡ Numbers of screening examinations among 31 271 (SFM) and 8380 (FFDM) women.§ Difference between FFDM and SFM in percentage of screening examinations that were routine recall or prevalent as opposedto referrals.
Table 2
Cancer Detection Rates, Recall Rates, and PPVs of an Abnormal Mammogram by Screening Modality
Age at Screening (y) SFM FFDMRelative Risk for FFDM vs SFM
Crude Crude P Value* Adjusted† Adjusted P Value*
Cancer Detection Rates per 100 Screening Mammograms (n � 40 198‡ and 34 755§ in crude and adjusted analyses, respectively)�60 0.52 (0.42, 0.62) 0.61 (0.40, 0.82) 1.18 (0.71, 1.64) .43 1.05 (0.59, 1.51) .83�60 0.84 (0.68, 1.00) 0.81 (0.50, 1.12) 0.97 (0.55, 1.38) .87 0.86 (0.48, 1.25) .52All ages 0.65 (0.56, 0.73) 0.68 (0.51, 0.86) 1.06 (0.75, 1.37) .70 0.95 (0.65, 1.25) .74
Recall Rates per 100 Screening Mammograms (n � 40 198‡ and 34 755§ in crude and adjusted analyses, respectively)�60 5.03 (4.72, 5.34) 5.23 (4.62, 5.83) 1.04 (0.90, 1.18) .56 0.93 (0.79, 1.06) .31�60 3.54 (3.22, 3.86) 4.08 (3.39, 4.76) 1.15 (0.93, 1.37) .15 1.01 (0.80, 1.23) .91All ages 4.43 (4.20, 4.65) 4.79 (4.33, 5.24) 1.08 (0.97, 1.20) .15 0.95 (0.84, 1.07) .44
PPV of Breast Cancers among Recalls (%) (n � 1810� and 1516# in crude and adjusted analyses, respectively)�60 10.29 (8.36, 12.22) 11.64 (7.85, 15.43) 1.13 (0.71, 1.56) .52 1.07 (0.65, 1.49) .72�60 23.67 (19.75, 27.59) 19.85 (13.02, 26.68) 0.84 (0.52, 1.16) .36 0.84 (0.51, 1.18) .39All ages 14.60 (12.75, 16.45) 14.29 (10.88, 17.69) 0.98 (0.71, 1.24) .87 0.95 (0.68, 1.23) .75
Note.—Numbers in parentheses are 95% CIs. Interactions between screening modality and age at screening were not statistically significant (P � .54, .49, and .37 for adjusted detection rates,recall rates, and PPVs, respectively).
* P value for comparison between SFM and FFDM.† Adjusted for age, ethnicity, type of referral, and area of residence.‡ 31 720 SFM and 8478 FFDM screening examinations.§ Smaller numbers due to missing information on ethnicity (27 598 SFM and 7157 FFDM screening examinations).� 1404 SFM and 406 FFDM positive screening mammograms.# Smaller numbers due to missing information on ethnicity (1182 SFM and 334 FFDM recalls).
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
Radiology: Volume 251: Number 2—May 2009 ▪ radiology.rsnajnls.org 349
mammogramsper year. The second readerwas aware of the first reader’s opinion; anydisagreementwas arbitrated independentlyby a third reader. These procedures fol-lowed NHSBSP guidelines (4) and were notspecifically implemented for the purpose ofour study.
Data routinely collected at the time ofeach screening examination included age,ethnicity, area of residence, type of refer-ral, screening round, and, for all malignan-cies, type (invasive vs noninvasive), size,andhistologic grade. For the purpose of ourstudy, the screening mammograms for alldetected breast cancer cases were re-evaluated by a radiologist (S.V.) blinded tothe findings at the original assessment andpathologic evaluation. The following wereassessed: mammographic density (BreastImaging Reporting and Data System [BI-RADS] classification [6]); location, type,and size of lesion; radiologic characteristicsof lesion; and degree of radiologic suspicion(R1 � normal; R2 � benign; R3 � atypical;R4 � suspicious; R5 � highly suggestive ofmalignancy). In UK and European practice,theR3 categoryprompts assessment ratherthan early recall (5).
Standard logistic regression models (7)were fitted to assess the association ofscreening modality with the probability ofdetecting breast cancer while adjusting forage, ethnicity, area of residence, and typeof referral. To account for within-womancorrelations, observations from a singlewoman were treated as a cluster, with 95%confidence intervals (CIs) computed by us-ing robust standard errors and Wald testsused to assess associations. Similar logisticregression models were used to comparerecall rates, positive predictive values(PPVs) of an abnormal mammogram, andbinary tumor characteristics between thetwo screening modalities (the latter re-stricted to cases). Tests for interactionwere performed to investigate whether theeffect of screening modality on cancer de-tection, recall, and PPV differed by age atscreening or screening round. Analyseswere conducted by using statistical soft-ware (Stata, version 10; Stata, College Sta-tion, Tex).
Systematic ReviewA computerized search was conducted byusing PubMed, MEDLINE, and EMBASE to
Table 3
Characteristics of Mammographically Detected Cancers by Screening Modality
Characteristic SFM FFDM P Value*
Mammograms with BI-RADS 3�
(heterogeneous or extremely dense) 64/205 (31.2) 15/58 (25.9) .31Mammographic lesion visible on both standard
views 165/201 (82.1) 46/54 (85.2) .86Location of mammographic lesion
Upper quadrants 127/181 (70.2) 35/51 (68.6) .98Lower quadrants 38/181 (21.0) 11/51 (21.6)Other (eg, upper and lower) 16/181 (8.8) 5/51 (9.8)
Type of mammographic lesionMass 120/201 (59.7) 32/54 (59.3) .97Calcification 73/201 (36.3) 16/54 (29.6) .27Architectural distortion 67/201 (33.3) 16/54 (29.6) .29Asymmetric density 32/201 (15.9) 7/54 (13.0) .94
Radiologic characteristics of mass lesionsMean size (mm)†‡ 15.8 (14.3, 17.2) 15.3 (12.6, 18.1) .59Ovoid shape 68/99 (68.7) 17/26 (65.4) .87Microlobulate/lobulate border 9/55 (16.4) 2/14 (14.3) .40Halo 1/120 (0.8) 2/32 (6.3) .10Multiple lesions 8/119 (6.7) 1/32 (3.1) .46
Radiologic characteristics of calcificationsPresent jointly with a mass 18/69 (26.1) 1/15 (6.7) .26Nature .67
Punctuate 23/70 (32.9) 6/16 (37.5)Variable 16/70 (22.9) 4/16 (25.0)Linear branching/pleomorphic 31/70 (44.2) 6/16 (37.5)
Distribution .57Cluster 26/71 (36.6) 7/15 (46.7)Linear 19/71 (26.8) 2/15 (13.3)Segmental/regional 24/71 (33.8) 6/15 (4.0)
Extent (mm)†§ 33.4 (24.6, 42.2) 31.0 (12.2, 49.8) .80Density of individual particles .16
High 32/72 (44.4) 10/16 (62.5)Moderate 20/72 (27.8) 5/16 (31.3)Low 20/72 (27.8) 1/16 (6.3)
Cluster conspicuity .12High 13/69 (18.8) 9/14 (64.3)Moderate 28/69 (40.6) 5/14 (35.7)Low 28/69 (40.6) 0/14
With high density in surrounding breast area 32/72 (44.4) 10/16 (62.5) .15Radiologic characteristics of architectural
distortionIsolated 6/67 (9.0) 3/16 (18.8) .16With mass/asymmetric density 6/67 (9.0) 1/16 (6.3) .89
Degree of malignancy suspicionR3 (atypical) 54/205 (26.3) 14/57 (24.6) .87R4 (suspicious) 68/205 (33.2) 17/57 (29.8)R5 (highly suggestive) 83/205 (4.5) 26/57 (45.6)
Histologic characteristicsInvasive 159/205 (77.6) 45/58 (77.6) .74Grade of invasive tumors .24
1 39/156 (25.0) 8/42 (19.1)2 78/156 (50.0) 18/42 (42.9)3 39/156 (25.0) 16/42 (38.1)
(Table 3 continues)
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
350 radiology.rsnajnls.org ▪ Radiology: Volume 251: Number 2—May 2009
identify studies published in English-lan-guage journals between January 1, 2000,andFebruary29, 2008 (inclusive) that com-pared FFDM to SFM in terms of their pro-cess indicators by using as search terms“digital” and “mammogra*” in the title.Studies were eligible if they reported esti-mates of detection rates, recall rates, orPPVs of an abnormal mammogram accord-ing to screening modality. Reference lists
within relevant articles and reviews weresearched to identify further publications.Each publication was checked to see if itwas a reanalysis or a subset of another eli-gible study and, if so, only the larger study,or constituent studies, where the data weredisplayedmost appropriatelywere includedto avoid duplication.
Each eligible publication was inde-pendently reviewed by two of the authors
(S.M.P.P., statistician; I.M.d.S.S., epide-miologist). Study-specific modality differ-ences (with 95% CIs) in process indica-tors were estimated if not provided in theoriginal publication; differences in PPVfrom paired studies were calculated byusing the methods outlined by Moskowitzand Pepe (8). Researchers from two stud-ies (9,10) were contacted to obtain addi-tional data, but additional data could nolonger be retrieved for one of these stud-ies (9). Forest plots were used to visualizestudy-specific estimates. Heterogeneitybetween studies was assessed by using a�2 statistic (Q) with inverse varianceweights. The percentage of total variationbetween studies that was caused by het-erogeneity rather than chance was quan-tified as I2 (11). Random effects modelswere used, if appropriate, to combinestudy-specific estimates (12) overall andstratified by study type and prespecifiedage groups (�60 and �60 years, or asclose to these as possible). Study size ef-fects were assessed by using funnel plotsand the Egger test (12).
Results
CELBSS StudyA total of 39 651womenunderwent 40 198screening examinations (546 women un-derwentmore than one screening examina-tion because of their moves between gen-eral practitioner practices). The medianage at screening was 58.0 years (inter-quartile range: 53.5, 63.5 years). Rela-tive to women in the SFM group, thosein the FFDM group were more likely tobe younger, Caucasian, and self-refer-rals (Table 1).
A total of 263 cancers were detectedin the study population (0.65 per 100screening mammograms). The crudedetection rate was slightly higher forFFDM than for SMF (0.68 vs 0.65 per100 screening mammograms, respec-tively), but after adjustment for age,ethnicity, type of referral, and area ofresidence there was no evidence of anydifference in detection rates betweenthe two screening modalities (Table 2).Similarly, there were no differences indetection rates between FFDM andSFM in analyses stratified by age at
Figure 1
Figure 1: Flow diagram illustrates the selection of studies included in the meta-analysis.
Table 3 (continued)
Characteristics of Mammographically Detected Cancers by Screening Modality
Characteristic SFM FFDM P Value*
Invasive tumor size (mm)†� 14.4 (12.8, 15.9) 17.01 (14.1, 19.9) .054
Note.—Except where noted, data are numbers of mammographically detected cancers with a given characteristic out ofrelevant SFM or FFDM tumors with no missing information, and numbers in parentheses are percentages.
* P value for difference in percentages between SFM and FFDM after adjustment for age, ethnicity, type of referral, and areaof residence.† Data are means, with 95% CIs in parentheses.‡ Among 120 (SFM) and 32 (FFDM) lesions measured.§ Among 73 (SFM) and 16 (FFDM) calcifications measured.� Among 159 (SFM) and 45 (FFDM) tumores measured.
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
Radiology: Volume 251: Number 2—May 2009 ▪ radiology.rsnajnls.org 351
Tabl
e4
Char
acte
rist
ics
ofSt
udie
sIn
clud
edin
the
Syst
emat
icRe
view
Stud
yLo
catio
n
Sour
cePo
pulat
ion
and
Elig
ible
Age
Grou
p*St
udy
Type
SFM
FFDM
Com
men
tsEq
uipm
enta
ndFi
lm
No.o
fW
omen
/No.
ofCa
ncer
sDi
gital
Syste
man
dVi
ewin
gM
odali
tyNo
.ofW
omen
/No
.ofC
ance
rs
Lew
inet
al,2
002
(18)
Colo
rado
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m;
elig
ible
ages
,�
40y
(mea
n,55
.6y)
Pros
pect
ive
paire
dde
sign
GEDM
RKo
dak
Min
-R20
00fil
m
6736
/33
†
GESe
nogr
aphe
2000
DPr
otop
type
two-
mon
itor
wor
ksta
tion
Soft-
copy
read
ing
6736
/27†
Two
stan
dard
view
sof
each
brea
stw
ithea
chte
chni
que;
inde
pend
entS
FMan
dFF
DMsi
ngle
read
ings
,with
acce
ssto
prio
rim
ages
and
hist
ory
deta
ils;p
ositi
veSF
Man
dFF
DMfu
rther
eval
uate
din
cons
ensu
sm
eetin
gsbe
twee
nth
etw
ora
diol
ogis
ts;f
ollo
w-u
p:1
year
Yam
ada
etal
,20
04(9
)M
iyag
i,Ja
pan
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m;
elig
ible
ages
,�
50y
Pros
pect
ive
paire
dde
sign
Siem
ens
Mam
mom
at30
0No
va(M
o/M
o,M
o/Rh
,Ts/
Mo)
Koda
kM
in-R
2000
film
480/
2‡
GESe
nogr
aphe
2000
D(M
o/M
o,Rh
/Rh,
Mo/
Rh)
Koda
kDr
yVie
w86
10la
ser
imag
eun
itHa
rd-c
opy
read
ing
480/
2‡On
em
edio
late
ralo
bliq
uevi
ewof
each
brea
stw
ithea
chte
chni
que;
inde
pend
entS
FMan
dFF
DMdo
uble
read
ings
;fol
low
-up:
1ye
ar
Skaa
neet
al,
2005
(Osl
oI)
(10)
Oslo
,No
rway
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m;
elig
ible
ages
,50
–69
y(m
ean,
58.2
y)
Pros
pect
ive
paire
dde
sign
Siem
ens
Mam
mom
at30
0w
ith29
kVM
o/M
oKo
dak
Min
-R20
00fil
mw
ithM
in-R
2190
scre
en
3683
/30
§
GESe
nogr
aphe
2000
Din
AOP
mod
eSo
ft-co
pyre
adin
g
3683
/26§
Two
stan
dard
view
sof
each
brea
stw
ithea
chte
chni
que;
inde
pend
entS
MF
and
FFDM
doub
lere
adin
g;se
para
teco
nsen
sus
mee
tings
forp
ositi
veSF
Man
dFF
DMst
udie
s(w
ithac
cess
topr
evio
usst
udie
s);f
ollo
w-u
p:2
year
s(in
clud
edin
terv
alca
ncer
san
dsc
reen
ing-
dete
cted
canc
ers
insu
bseq
uent
roun
d)Pi
sano
etal
,200
5(D
MIS
T)(1
9,23
)
34Ce
nter
sin
U.S.
and
Cana
da
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
ms;
elig
ible
ages
NA(m
ean,
54.9
y;IQ
R:47
,62)
Pros
pect
ive
paire
dde
sign
;ran
dom
allo
catio
nof
mod
ality
orde
r
Syst
emro
utin
ely
used
inea
chce
nter
�
4276
0/17
4Fi
vedi
ffere
ntty
pes
ofdi
gita
lm
amm
ogra
phy
with
soft-
and/
orha
rd-c
opy
read
ing�
4276
0/18
5Tw
ost
anda
rdvi
ews
ofea
chbr
east
with
each
tech
niqu
e;in
depe
nden
tSFM
and
FFDM
sing
lere
adin
gs;F
FDM
:co
mbi
ned
data
fort
hefiv
esy
stem
s,as
they
show
edsi
mila
rfind
ings
;fol
low
-up:
455
days
(als
ow
ithan
alys
isat
1ye
ar)
Skaa
neet
al,
2007
(Osl
oII)
(20)
Oslo
,No
rway
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m;
elig
ible
ages
,45
–69
y(4
2%w
ere
aged
45–4
9y)
Rand
omiz
edtri
al;
rand
omal
loca
tion
stra
tified
byag
e
Siem
ens
Mam
mom
at30
0Ko
dak
Min
-R20
00fil
man
dM
in-R
2190
scre
ens
Ages
45–4
9y:
7082
/14
Ages
50–6
9y:
9903
/50
GESe
nogr
aphe
2000
DSo
ft-co
pyre
adin
g
Ages
45–4
9y:
2935
/9Ag
es 50–6
9y:
4009
/32
Two
stan
dard
view
sof
each
brea
stw
ithea
chte
chni
que;
inde
pend
entS
FMan
dFF
DMdo
uble
read
ings
;all
read
ers
have
take
npa
rtin
Oslo
Istu
dy;p
riori
mag
esno
tava
ilabl
eat
initi
alre
adin
gbu
tava
ilabl
ein
cons
ensu
sm
eetin
gs;f
ollo
w-u
p:2
year
sat
ages
50–6
9y,
1.5
year
atag
es45
-49
y(in
clud
edin
terv
alca
ncer
san
dsc
reen
ing-
dete
cted
canc
ers
insu
bseq
uent
roun
d)
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
352 radiology.rsnajnls.org ▪ Radiology: Volume 251: Number 2—May 2009
DelT
urco
etal
,20
07(2
1)Fl
oren
ce,
Italy
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m;
elig
ible
ages
,50–
69y
(47%
aged
50–5
9y)
Two
conc
urre
ntre
trosp
ectiv
eco
horts
,m
atch
edby
age
and
repo
rting
radi
olog
ist;
allo
catio
nto
scre
enin
gm
odal
itym
ainl
yby
resi
denc
e
Siem
ens
Mam
mom
at10
0014
385/
84GE
Seno
grap
he20
00D
Mam
moR
epor
tw
orks
tatio
nSo
ft-co
pyre
adin
g
1438
5/10
4Vi
ews
perb
reas
tnot
avai
labl
e;SF
Man
dFF
DMdo
uble
read
ings
butn
otin
depe
nden
t;re
call
base
don
susp
icio
nby
eith
erre
ader
(sin
gle-
read
erre
call)
;prio
rim
ages
avai
labl
eat
initi
alre
adin
g;fo
llow
-up:
scre
enin
g-de
tect
edca
ncer
son
ly
Vige
land
etal
,20
08(2
2)Ve
stfo
ldCo
unty
,No
rway
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m(fi
rst
prev
alen
ttw
o-ye
arro
und)
;ag
es50
–69
y(m
ean:
59y)
FFDM
popu
latio
n:w
omen
who
atte
nded
first
prev
alen
trou
ndin
Vest
fold
Coun
ty;c
ontro
lpo
pula
tion:
hist
oric
data
base
ofSF
Mpr
eval
ent
two-
year
roun
din
othe
r18
coun
ties
Siem
ens
Mam
mom
at30
0/30
00/3
000
Nova
GESe
nogr
aphe
DMR,
Seno
grap
he80
0T
Inst
rum
enta
rium
(Alp
ha/D
iam
ond)
Koda
kfil
ms
(Min
-RE
,Min
-R20
00)
324
763/
2105
Two
stat
icLo
rad
Sele
nia
units
(Hol
ogic
)Fl
at-p
anel
sele
nium
dete
ctor
sw
ith70
-�m
pixe
lsiz
eSo
ft-co
pyre
adin
g
1823
9/14
0Tw
ost
anda
rdvi
ews
ofea
chbr
east
with
each
tech
niqu
e;FF
DMan
dSM
Fin
depe
nden
tdou
ble
read
ing;
cons
ensu
sm
eetin
gsw
ithac
cess
topr
ior
mam
mog
ram
sfo
rFFD
Mpo
pula
tion,
varia
ble
for
cont
rolp
opul
atio
n;fo
llow
-up:
scre
enin
g-de
tect
edca
ncer
son
ly
Vinn
icom
beet
al,
2008
East Lo
ndon
,En
glan
d
Popu
latio
n-ba
sed
scre
enin
gpr
ogra
m;
ages
�50
y(m
edia
n,58
y;IQ
R:54
,64)
Retro
spec
tive
coho
rt;al
loca
tion
tosc
reen
ing
mod
ality
mai
nly
byar
eaof
resi
denc
e
Two
mob
ileLo
rad
MIV
units
3171
8/20
5#
Thre
eGE
Seno
grap
heDS
units
One
Lora
dSe
leni
aun
it
8475
/58#
Two
stan
dard
view
sof
each
brea
stw
ithea
chte
chni
que;
SFM
and
FFDM
doub
lere
adin
gsbu
tnot
inde
pend
ently
(dis
agre
emen
tsar
bitra
ted
byth
irdre
ader
);fo
llow
-up:
scre
enin
g-de
tect
edca
ncer
son
ly
*IQ
R�
inte
rqua
rtile
rang
e,NA
�no
tav
aila
ble.
†67
36sc
reen
ing
mam
mog
ram
sam
ong
4489
wom
en.
‡No
scre
enin
g-de
tect
edca
ncer
sha
doc
curr
edup
toen
dof
follo
w-u
p(th
etw
oca
ncer
sar
edi
agno
stic
case
sad
ded
toth
est
udy)
.§
Incl
udes
10in
terv
alca
ncer
san
d16
scre
enin
g-de
tect
edca
ncer
sin
subs
eque
ntro
und
(can
cers
diag
nose
dat
the
initi
alro
und:
28at
SFM
,23
atFF
DM).
�SF
Msy
stem
s:Ag
faHD
RC;S
terli
ngCT
/Kod
akM
inR-
2000
;Fuj
iMed
ical
Syst
ems,
Koda
kM
ini-R
2000
.FFD
Msy
stem
s:Se
noSc
an(F
isch
erM
edic
al);
Com
pute
dRa
diog
raph
ySy
stem
forM
amm
ogra
phy
(Fuj
iMed
ical
);Se
nogr
aphe
2000
D(G
EM
edic
alSy
stem
s);D
igita
lMam
mog
raph
ySy
stem
(Hol
ogic
);an
dSe
leni
aFu
llFi
eld
Digi
talM
amm
ogra
phy
Syst
em(H
olog
ic).
#31
718
scre
enin
gm
amm
ogra
ms
amon
g31
385
wom
enat
SFM
and
8475
scre
enin
gm
amm
ogra
ms
amon
g83
77w
omen
atFF
DM.
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
Radiology: Volume 251: Number 2—May 2009 ▪ radiology.rsnajnls.org 353
screening and no evidence that thesecomparisons varied by age (P � .54 forinteraction; Table 2) or screening round(ie, incident vs prevalent; P � .60 forinteraction).
The overall recall rate was 4.5% anddid not differ by screening modality, ei-
ther before or after adjustment for po-tential confounders (Table 2). Therewas also no evidence that the effect ofscreening modality on recall rates var-ied according to age at screening (P �.49 for interaction).
In total, 1810 women were recalled
for further assessment and 263 cancerswere detected and histologically con-firmed, yielding an overall PPV for anabnormal screening mammogram of14.5%. There was no difference in PPVby screening modality, either before orafter adjustment for potential confound-
Figure 2
Figure 2: Detection rates by screening modality. (a) Study-specific detection rates forSFM (black) and FFDM (gray), overall and by age at screening (when available), by studydesign. PM � pre- or perimenopausal. (b) Study-specific and pooled modality differences indetection rates, by study design. (c) Study-specific and pooled modality differences in detec-tion rates, by age at screening. No estimates were available for Yamada et al (9), as no cancerswere diagnosed. For Skaane et al (10), 2005, cancer detection rates are based on initial read-ing sessions and number of cancers diagnosed in the initial round, interval, and subsequentscreening round. For Pisano et al (19,23) data are also presented for certain population sub-groups defined by age, menopausal status, and breast density. Area of black squares in b andc is inversely proportional to the variance (on the log scale).
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
354 radiology.rsnajnls.org ▪ Radiology: Volume 251: Number 2—May 2009
ers (Table 2). The effect of screeningmodality on PPVs did not differ by ageat screening (P � .37 for interaction).
Review of the original screeningmammograms for the 263 detected can-cers showed no associations betweenscreening modality and BI-RADS mam-mographic density (Table 3). Therewere also no associations betweenscreening modality and type of mammo-graphic lesion, likelihood of it being vis-ible on both standard views, its location,or radiologic characteristics. There wasno difference between the two modalitiesin the proportion of detected tumors thatwere invasive or in the distribution oftheir histologic grades. The average inva-sive tumor size was slightly larger for tu-mors detected with FFDM, but after ad-justment for potential confounders, thedifference was only of borderline statisti-cal significance (difference of 3.45 mm[95% CI: �0.07 mm, 6.97 mm]; P �.054).
Systematic ReviewThe search identified 189 publications(Fig 1); 124 were on digital mammogra-phy only and 52 were not eligible. Of theremaining 13 publications, one was on astudy performed in a diagnostic popula-tion (13) and in four (14–17), the studypopulation overlapped with that inother eligible publications. Thus, datafrom seven studies extracted from eightpublications (9,10,18–23), along withour UK study, were included in the re-view (Fig 1; Table 4). Four studies(9,10,18,19) had a paired design, thatis, the same group of women underwentscreening with both modalities, onestudy (20) was a randomized controlledtrial, and two studies (21,22)—plus ourstudy—had a cohort design (althoughsome included only baseline data intheir analyses). All studies, except one(9), used two standard views of eachbreast with each screening modality.Only our study and two others (9,19)used FFDM with hard-copy image read-ing, although this was only partial forone of the studies (19) (Table 4). Twostudies (18,19) used single reading, fivestudies (9,10,20,22), including ours,used double reading (three with consen-sus [10,20,22] and one with arbitration
by a third reader [our study]), and onestudy (21) used double reading with sin-gle-reader recall; however, in the latterstudy (21), as in our study, the tworeadings were not independent. In somestudies, prior mammograms and historydetails were available either at the ini-tial reading (18,21) or at the consensusmeetings (10,20,22). The length of fol-low-up varied, with two studies (10,20)including interval cancers as well as can-cers detected in the subsequent screen-ing round, while some cohort studiesincluded only baseline data.
Figure 2a shows study-specific de-tection rates for SFM and FFDM at allages, and by age group whenever avail-able in the original publication. Detec-tion rates tended to be lower in the twoU.S. studies and, as expected, in theyounger age groups. Figure 2b shows
study-specific and pooled modality dif-ferences in detection rates; there wasno evidence of study size effects (Eggerfunnel plot asymmetry test, P � .85).The overall pooled estimate was consis-tent with FFDM having a higher detec-tion rate than SFM (pooled FFDM-SFMdifference, 0.04 [95% CI: �0.03, 0.11]per 100 screening mammograms, equiva-lent to FFDM depicting an extra fourcases of breast cancer per every 10 000screening mammograms), but withevidence of some heterogeneity be-tween study types (I2 � 40%) (Fig 2b).Analyses stratified by age showedhigher FFDM detection rates at 60years or younger (equivalent to FFDMdepicting an extra 11 [95% CI: 4, 18]breast cancer cases per every 10 000screening mammograms) with no be-tween-study type heterogeneity (I2 �
Figure 3
Figure 3: Study-specific recall rates by screening modality (black � SFM, gray � FFDM) by study de-sign. For Pisano et al (19), recall rates are those reported for the BI-RADS scale.
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
Radiology: Volume 251: Number 2—May 2009 ▪ radiology.rsnajnls.org 355
0%) (Fig 2c). Pooled estimates at olderages were more difficult to interpret ow-ing to fewer and smaller studies. Theonly study (19,23) to have comparedthe two modalities simultaneously byage, menopausal status, and breast den-sity did not find any differences exceptthat FFDM yielded higher detectionrates among pre- and perimenopausalwomen younger than 50 years withdense breasts, whereas SFM yieldedslightly higher detection rates amongwomen older than 64 years with fatty(nondense) breasts (Fig 2a).
Recall rates varied greatly between
studies, with much higher rates in theUnited States than in the European andJapanese studies (Fig 3). There wasmarkedbetween-studyheterogeneity in dif-ferences in recall rates between modalities(I2 � 94%), with some studies showing sig-nificantly lower and others significantlyhigher recall rates for FFDM; thus, pooledestimates could not be calculated. Similarly,there was marked between-study type het-erogeneity in modality differences in thePPV of an abnormal mammogram (I2 �100%), with only cohort studies showing ahigher PPV for FFDM (Fig 4) and, hence,no pooled estimates were calculated. Some
studies (9,18–21) presented various recalland PPV estimates for different definitionsof abnormal mammograms (eg, before orafter consensus meetings) and detectedcancers (eg, at initial screening only or dur-ing follow-up), but these alternative esti-mates did not affect the findings shown inFigures 3 and 4.
Discussion
We found similar cancer detection ratesfor SFM and FFDM using hard-copy im-age reading; meta-analysis of publisheddata, along with our UK study findings,showed that FFDM performed, in termsof process measures, at least as well asconventional SFM.
Only two other published studies toour knowledge have compared FFDMwith SFM within the context of the dailypractice of a routine screening program(21,22). Del Turco et al (21) comparedFFDM and SFM within a local popula-tion-based program in Italy and found aslightly higher cancer detection rate forFFDM, which was particularly markedfor cancers that manifested as clusteredcalcifications (0.26% for FFDM vs0.12% for SFM; P � .007). Vigelandet al (22) conducted their study withinthe Norwegian national screening pro-gram and reported a higher detectionrate for FFDM, which was particularlymarked for ductal carcinoma in situ(0.21% for FFDM vs 0.11% for SFM;P � .001).
To our knowledge, our study is theonly one to have examined cancer de-tection rates for FFDM by using hard-copy image reading. Although Yamadaet al (9) used hard-copy reading, detec-tion rates could not be estimated, be-cause no cancers occurred during thefollow-up period. Pisano et al (19) alsoused hard-copy images for some of theirFFDM systems but these were not ana-lyzed separately. The use of hard-copyimage reading in our study was a matterof necessity rather than choice, becauseat the time, FFDM was still undergoingevaluation by the NHSBSP, and thusFFDM with soft-copy reading had notbeen incorporated into routine screen-ing. It did, however, obviate some of thedifficulties reported by authors of
Figure 4
Figure 4: Study-specific PPVs of an abnormal mammogram by screening modality (black � SFM, gray �FFDM), by study design and age. For Skaane et al (10), 2005, PPVs are based on all cancers diagnosed ininitial round, interval, and subsequent screening round among those with SFM or FFDM positive screeningmammograms at initial reading (prior to consensus meetings). For Pisano et al (19,23), PPVs are those re-ported for the BI-RADS scale. For Skaane et al (20), 2007, PPVs are based on number of cancers diagnosed ininitial round, interval, and subsequent screening round among those recalled (after consensus meetings). ForPisano et al (19,23) data are also presented for certain population subgroups defined by age, menopausalstatus (PM � pre- or perimenopausal), and breast density.
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
356 radiology.rsnajnls.org ▪ Radiology: Volume 251: Number 2—May 2009
other studies, such as use of prototypesoft-copy workstations (18), subopti-mal viewing conditions, and learningcurve effect (15). In our study allscreening mammograms, whether fromSFM or FFDM, were viewed with iden-tical conditions. For the first 18 monthsof the study, it was not possible to printhard copy of the optimally postpro-cessed GE Healthcare images (“Pre-mium View”), and FFDM screening im-ages obtained during this period resem-bled analogue film images. Thus, somecancers may have been missed at FFDMin this group. This did not apply to thehard-copy images from screening mam-mograms obtained with the Selenia unit(Lorad), which were indistinguishablefrom the soft-copy images. Very fewstudies have compared hard-copy andsoft-copy reading, but in one analysis of60 tumors (31 benign, 29 malignant),there was little difference in sensitivitybetween the two modes, although threeof the four readers improved their spec-ificity and PPV with soft-copy imagereading (24). If true, these findingswould suggest that hard-copy readingmight not have affected the observedcancer detection rates for FFDM but itmight have decreased the specificity andPPV of this screening modality, thusleading to an increase in its recall rate(15,19).
The overall recall rate of 4.5% inour study was well below the NHSBSPtarget of less than 7% for prevalentscreening examinations and 5% for inci-dent screening examinations (4). Recallrates were significantly higher in theFFDM group than the SFM group in two(20,21) of the three studies (20–22)that found higher FFDM detectionrates, whereas recall rates in our studywere very similar for the two modali-ties. It is possible that the lack of FFDM-SFM difference in detection rates in ourstudy may be partly due to the fact thatrecall rates were similar for the two mo-dalities. Skaane et al (20) commented intheir article that their SFM recall ratewas too low; this might have led to anoverestimation of the FFDM-SFM dif-ference in detection rates in their study.
Meta-analyses provide an opportu-nity to identify sources of between-
study heterogeneity and, when appro-priate, to combine study-specific effectsto estimate an overall summary effectacross the various studies. Our meta-analysis was consistent with FFDM per-forming at least as well as conventionalSFM. Subgroup analyses by age pro-vided some evidence of FFDM having aslightly higher detection rate than SFMat age 60 years or younger, equivalentto an extra 11 breast cancer cases de-tected at FFDM per every 10 000screening examinations, but this findingneeds to be confirmed in future studies.Cohort studies, such as ours, are moreprone to be affected by confoundingthan paired studies or large randomizedtrials. In cohort studies, distinct groupsof women underwent FFDM and SFM,with the allocation to screening modal-ity being determined nonrandomly bythe women themselves or by the screen-ing program. Thus, women who under-went FFDM might have differed fromthose who underwent SFM in relation tofactors that influence detection rates.Although attempts were made to mini-mize confounding at the design andanalysis stages, one cannot exclude thepossibility that the findings from cohortstudies might have been affected by re-sidual or unknown confounding. Reas-suringly, the meta-analyses found noclear evidence of heterogeneity by studydesign, particularly in analyses stratifiedby age. Although meta-analyses areprone to small size effects, includingpublication bias, the data did not pro-vide any evidence to support this.
In summary, the findings from ourUK study and the meta-analysis are con-sistent with FFDM performing at leastas well as conventional SFM for breastcancer screening. Further research is,however, required to assess the eco-nomic costs of FFDM and identify thebest clinical and cost-effective ways ofimplementing it. �
References1. Humphrey LL, Helfand M, Chan BKS, Woolf
SH. Breast cancer screening: a summary ofthe evidence for the US Preventive ServicesTask Force. Ann Intern Med 2002;137:347–360.
2. Fletcher SW, Elmore JG. Clinical practice:
mammographic screening for breast cancer.N Engl J Med 2003;348:1672–1680.
3. Dershaw DD. Status of mammography afterthe Digital Mammography Imaging Screen-ing Trial: digital versus film. Breast J 2006;12:99–102.
4. Liston J, Wilson R, eds. Quality assuranceguidelines for breast cancer screening radiol-ogy. National Health Service Breast Screen-ing Programme Publication No 59. London,England: Department of Health, 2005.
5. Perry N, Broeders M, de Wolf C, et al, eds.European guidelines for quality assurance inbreast cancer screening and diagnosis. 4thed. Luxembourg: European Communities,2006.
6. American College of Radiology. BI-RADSmammography breast imaging atlas. 4th ed.Reston, Va: American College of Radiology,2003.
7. Clayton D, Hills M. Statistical models in epi-demiology. Oxford, England: Oxford Univer-sity Press, 1993.
8. Moskowitz CS, Pepe MS. Comparing thepredictive values of diagnostic tests: samplesize and analyses for paired study designs.Clin Trials 2006;3:272–279.
9. Yamada T, Saito M, Ishibashi T, et al. Com-parison of screen-film and full-field digitalmammography in Japanese population-basedscreening. Radiat Med 2004;22:408–412.
10. Skaane P, Skjennald A, Young K, et al. Fol-low-up and final results of the Oslo I studycomparing screen-film mammography andfull-field digital mammography with soft-copy reading. Acta Radiol 2005;46:679–689.
11. Higgins JP, Thompson SG, Deeks JJ, AltmanDG. Measuring inconsistency in meta-analy-ses. BMJ 2003;327:557–560.
12. Sutton AJ. Methods for meta-analysis inmedical research. Wiley Series in Probabil-ity and Statistics. New York, NY: Wiley,2000.
13. Venta LA, Hendrick RE, Adler YT, et al.Rates and causes of disagreement in inter-pretation of full-field digital mammographyand film-screen mammography in a diagnos-tic setting. AJR Am J Roentgenol 2001;176:1241–1248.
14. Lewin JM, Hendrik RE, D’Orsi CJ, et al.Comparison of full-field digital mammogra-phy with screen-film mammography for can-cer detection: results of 4,945 paired exami-nations. Radiology 2001;218:873–880.
15. Skaane P, Young K, Skjennald A. Popula-tion-based mammography screening: com-parison of screen-film and full-field digital
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
Radiology: Volume 251: Number 2—May 2009 ▪ radiology.rsnajnls.org 357
mammography with soft-copy reading—OsloI Study. Radiology 2003;229:877–884.
16. Skaane P, Hofvind S, Skennald A. Screen-film mammography versus full-field digitalmammography with soft-copy reading: ran-domised trial in a population-based screen-ing program—the Oslo II Study. Radiology2004;232:197–204.
17. Skaane P, Balleyguier C, Diekmann F, et al.Breast lesion detection and classification:comparison of screen-film mammographyand full-field digital mammography with soft-copy reading-observer performance study.Radiology 2005;237:37–44.
18. Lewin JM, D’Orsi CJ, Hendrick RE, et al.Clinical comparison of full-field digital mam-
mography and screen-film mammographyfor detection of breast cancer. AJR Am JRoentgenol 2002;179:671–677.
19. Pisano ED, Gatsonis C, Hendrick E, et al.Diagnostic performance of digital versus filmmammography for breast cancer screening.N Engl J Med 2005;353:1773–1783.
20. Skaane P, Hofvind S, Skjennald A. Ran-domised trial of screen-film versus full-fielddigital mammography with soft-copy read-ing in population-based screening pro-gram: follow-up and final results of Oslo IIstudy. Radiology 2007;244:708–717.
21. Del Turco MR, Mantellini P, Ciatto S, et al.Full-field digital versus screen-film mam-mography in concurrent screening co-
horts. AJR Am J Roentgenol 2007;189:860–866.
22. Vigeland E, Klaasen H, Klingen TA, HofvindS, Skaane P. Full-field digital mammographycompared with screen film mammography inthe prevalent round of a population-basedscreening program: the Vestfold Countystudy. Eur Radiol 2008;18:183–191.
23. Pisano ED, Hendrick RE, Yaffe MJ, et al.Diagnostic accuracy of digital versus filmmammography: exploratory analysis of se-lected population subgroups in DMIST. Radi-ology 2008;246:376–383.
24. Obenauer S, Hermann K-P, Marten K, et al.Soft copy versus hard copy reading in digitalmammography. J Digit Imaging 2003;16:341–344.
BREAST IMAGING: Full-Field Digital versus Screen-Film Mammography Vinnicombe et al
358 radiology.rsnajnls.org ▪ Radiology: Volume 251: Number 2—May 2009
Radiology 2009This is your reprint order form or pro forma invoice
(Please keep a copy of this document for your records.)
Author Name _______________________________________________________________________________________________Title of Article _______________________________________________________________________________________________Issue of Journal_______________________________ Reprint # _____________ Publication Date ________________Number of Pages_______________________________ KB # _____________ Symbol RadiologyColor in Article? Yes / No (Please Circle)Please include the journal name and reprint number or manuscript number on your purchase order or other correspondence.
Order and Shipping Information
Reprint Costs (Please see page 2 of 2 for reprint costs/fees.)
________ Number of reprints ordered $_________
________ Number of color reprints ordered $_________
________ Number of covers ordered $_________
Subtotal $_________
Taxes $_________(Add appropriate sales tax for Virginia, Maryland, Pennsylvania, and the District of Columbia or Canadian GST to the reprints if your order is to be shipped to these locations.)
First address included, add $32 for each additional shipping address $_________
TOTAL $_________
Shipping Address (cannot ship to a P.O. Box) Please Print Clearly
Name ___________________________________________Institution _________________________________________Street ___________________________________________City ____________________ State _____ Zip ___________Country ___________________________________________Quantity___________________ Fax ___________________Phone: Day _________________ Evening _______________E-mail Address _____________________________________
Additional Shipping Address* (cannot ship to a P.O. Box)
Name ___________________________________________Institution _________________________________________Street ___________________________________________City ________________ State ______ Zip ___________Country _________________________________________Quantity __________________ Fax __________________Phone: Day ________________ Evening ______________E-mail Address ____________________________________* Add $32 for each additional shipping address
Payment and Credit Card DetailsEnclosed: Personal Check ___________
Credit Card Payment Details _________Checks must be paid in U.S. dollars and drawn on a U.S. Bank.Credit Card: __ VISA __ Am. Exp. __ MasterCardCard Number __________________________________Expiration Date_________________________________Signature: _____________________________________
Please send your order form and prepayment made payable to:
Cadmus ReprintsP.O. Box 751903Charlotte, NC 28275-1903
Note: Do not send express packages to this location, PO Box.FEIN #:541274108
Invoice or Credit Card InformationInvoice Address Please Print ClearlyPlease complete Invoice address as it appears on credit card statementName ____________________________________________Institution ________________________________________Department _______________________________________Street ____________________________________________City ________________________ State _____ Zip _______Country ___________________________________________Phone _____________________ Fax _________________E-mail Address _____________________________________
Cadmus will process credit cards and Cadmus Journal Services will appear on the credit card statement.
If you don’t mail your order form, you may fax it to 410-820-9765 with your credit card information.
Signature __________________________________________ Date _______________________________________Signature is required. By signing this form, the author agrees to accept the responsibility for the payment of reprints and/or all charges described in this document.
Reprint order forms and purchase orders or prepayments must be received 72 hours after receipt of form either by mail or by fax at 410-820-9765. It is the policy of Cadmus Reprints to issue one invoice per order.
Please print clearly.
Page 1 of 2RB-1/01/09
Radiology 2009Black and White Reprint Prices
Domestic (USA only)# of
Pages50 100 200 300 400 500
1-4 $239 $260 $285 $303 $323 $3405-8 $379 $420 $455 $491 $534 $5729-12 $507 $560 $651 $684 $748 $814
13-16 $627 $698 $784 $868 $954 $1,03817-20 $755 $845 $947 $1,064 $1,166 $1,27221-24 $878 $985 $1,115 $1,250 $1,377 $1,51825-28 $1,003 $1,136 $1,294 $1,446 $1,607 $1,75729-32 $1,128 $1,281 $1,459 $1,632 $1,819 $2,002
Covers $149 $164 $219 $275 $335 $393
International (includes Canada and Mexico)# of
Pages50 100 200 300 400 500
1-4 $299 $314 $367 $429 $484 $5465-8 $470 $502 $616 $722 $838 $9499-12 $637 $687 $852 $1,031 $1,190 $1,369
13-16 $794 $861 $1,088 $1,313 $1,540 $1,76517-20 $963 $1,051 $1,324 $1,619 $1,892 $2,16821-24 $1,114 $1,222 $1,560 $1,906 $2,244 $2,58825-28 $1,287 $1,412 $1,801 $2,198 $2,607 $2,99829-32 $1,441 $1,586 $2,045 $2,499 $2,959 $3,418
Covers $211 $224 $324 $444 $558 $672
Minimum order is 50 copies. For orders larger than 500 copies, please consult Cadmus Reprints at 800-407-9190.
Reprint CoverCover prices are listed above. The cover will include the publication title, article title, and author name in black.
ShippingShipping costs are included in the reprint prices. Do mesticorders are shipped via FedEx Ground service. Foreign orders are shipped via a proof of delivery air service.
Multiple ShipmentsOrders can be shipped to more than one location. Please be aware that it will cost $32 for each additional location.
DeliveryYour order will be shipped within 2 weeks of the journal print date. Allow extra time for delivery.
Color Reprint PricesDomestic (USA only)
# of Pages
50 100 200 300 400 500
1-4 $247 $267 $385 $515 $650 $7805-8 $297 $435 $655 $923 $1194 $14679-12 $445 $563 $926 $1,339 $1,748 $2,162
13-16 $587 $710 $1,201 $1,748 $2,297 $2,84317-20 $738 $858 $1,474 $2,167 $2,846 $3,53221-24 $888 $1,005 $1,750 $2,575 $3,400 $4,23025-28 $1,035 $1,164 $2,034 $2,986 $3,957 $4,91229-32 $1,186 $1,311 $2,302 $3,402 $4,509 $5,612
Covers $149 $164 $219 $275 $335 $393
International (includes Canada and Mexico))# of
Pages50 100 200 300 400 500
1-4 $306 $321 $467 $642 $811 $9865-8 $387 $517 $816 $1,154 $1,498 $1,8449-12 $574 $689 $1,157 $1,686 $2,190 $2,717
13-16 $754 $874 $1,506 $2,193 $2,883 $3,57017-20 $710 $1,063 $1,852 $2,722 $3,572 $4,42821-24 $1,124 $1,242 $2,195 $3,231 $4,267 $5,30025-28 $1,320 $1,440 $2,541 $3,738 $4,957 $6,15329-32 $1,498 $1,616 $2,888 $4,269 $5,649 $7028
Covers $211 $224 $324 $444 $558 $672
Tax DueResidents of Virginia, Maryland, Pennsylvania, and the District of Columbia are required to add the appropriate sales tax to each reprint order. For orders shipped to Canada, please add 7% Canadian GST unless exemption is claimed.
OrderingReprint order forms and purchase order or prepayment is required to process your order. Please reference journal name and reprint number or manuscript number on anycorrespondence. You may use the reverse side of this form as a proforma invoice. Please return your order form and prepayment to:
Cadmus ReprintsP.O. Box 751903Charlotte, NC 28275-1903
Note: Do not send express packages to this location, PO Box.FEIN #:541274108
Please direct all inquiries to:
Rose A. Baynard800-407-9190 (toll free number)410-819-3966 (direct number)410-820-9765 (FAX number)[email protected] (e-mail)
Reprint Order Forms and purchase order or prepayments must be received 72 hours after receipt of form.
Page 2 of 2