679

Rate of Insufficient Samples for Fine-NeedleAspiration for Nonpalpable Breast Lesions in aMulticenter Clinical TrialThe Radiologic Diagnostic Oncology Group 5 Study

BACKGROUND. Radiologic Diagnostic Oncology Group 5 is a multicenter clinicalEtta D. Pisano, M.D.1

Laurie L. Fajardo, M.D.2 trial designed to evaluate fine-needle aspiration (FNA) of nonpalpable breast le-

sions performed by multiple operators using the same protocol.John Tsimikas, Ph.D.3

Nour Sneige, M.D.4 METHODS. Four hundred and nineteen women with mammographically detected

nonpalpable breast lesions were enrolled on the trial at 18 institutions. Group AWilliam J. Frable, M.D.5

Constantine A. Gatsonis, Ph.D.6 institutions randomized women to stereotactically guided FNA (SFNA) followed

by stereotactically guided core needle biopsy (SCNB), or SCNB only. Group BW. Phil Evans, M.D.7

Irena Tocino, M.D.8 institutions randomized women to SFNA and SCNB, SCNB, or ultrasonographically

guided FNA followed by ultrasonographically guided core needle biopsy (USCNB),Barbara J. McNeil, M.D., Ph.D.3 for TheRDOG5 Investigators or USCNB only. A total of 377 women were eligible for analysis.

RESULTS. FNA yielded 128 insufficient samples for the 377 patients (33.95%; 95%1 Department of Radiology, Lineberger Compre- confidence interval, 29.2–38.7%). The rate of insufficient samples varied by type ofhensive Cancer Center, University of North Car- lesion with calcified lesions associated with a significantly higher rate of insufficientolina-Chapel Hill, Chapel Hill, North Carolina. sampling than masses (P õ 0.001). The radiologist’s level of suspicion of the lesion2 Department of Radiology, University of Vir- was not a statistically significant predictor of insufficient samples for mass lesions,ginia, Health Sciences Center, Charlottesville, but was a predictor for calcified lesions. For the 336 lesions for which histologicVirginia. information was available, insufficient samples occurred in significantly more be-3 Department of Health Care Policy, Harvard nign than malignant lesions.University, Boston, Massachusetts. CONCLUSIONS. The high rate of insufficient samples for FNA of nonpalpable breast

lesions in this multicenter trial makes its use impractical in this setting. Because4 Department of Pathology, University of Texas,M. D. Anderson Cancer Center, Houston, Texas. of this factor, the study was terminated early. Cancer 1998;82:679–88.

q 1998 American Cancer Society.5 Department of Pathology, Virginia Common-wealth University, Richmond, Virginia.

KEYWORDS: fine-needle aspiration, rate of insufficient samples, nonpalpable breast6 Center for Statistical Sciences, Brown Univer-lesions, multicenter clinical trial.sity, Providence, Rhode Island.

7 Department of Radiology, Baylor-Komen

ABreast Center, Dallas, Texas. lthough screening mammography in conjunction with breastphysical examination is believed to reduce breast carcinoma mor-

8 Department of Radiology, Yale University, New tality for women age ú 50 years, large numbers of breast biopsies forHaven, Connecticut.

Wendie A. Berg, Denise Farleigh, John Milbrath, Bonnie Yankaskas.This work was funded by National Institutes ofHealth grants UO1 CA62476, UO1 CA62514, Stephen Feig, Murray Rebner, Daniel Sullivan,

Frank Bradley, M. Patricia Braeuning, James Address for reprints: Etta D. Pisano, M.D., CBand UO1 CA62462.Connelly, David Page, and Stuart Schnitt. The 7510 Room 503, Old Infirmary Building, Univer-following individuals also made substantial con- sity of North Carolina-Chapel Hill, Chapel Hill,The authors gratefully acknowledge the contri-

butions of the following people in the comple- tributions: Mary Baldwin, Thomas Caldwell, NC 27599-7510.Daryl Caudry, Barbara Croft, Kristen Harris,tion of this work: The Radiologic Diagnostic On-

cology Group 5 investigators: Steven Parker, Brenda Harrison, Eva Hoke, Dana Jeffries, Cyn- Received June 6, 1997; revision received Sep-thia Olson, Elaine Pakuris, Joanne Stetz, and tember 2, 1997; accepted September 2, 1997.Gillian Newstead, Debra Monticciolo, Jana Rice,

q 1998 American Cancer Society

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

680 CANCER February 15, 1998 / Volume 82 / Number 4

benign lesions are generated as a result of these detec- samples and atypical or benign cytologies are consid-ered negative, the reported sensitivity of FNA cytologytion methods.1–3 Only 10–30% of nonpalpable lesions

subjected to open surgical biopsy prove to be malig- has varied from 43.8–91.7%, the specificity has rangedfrom 89.8–100%, the predictive value of a positive ex-nant tumors.4–8 Generally speaking, the morbidity

from open biopsy of nonpalpable benign lesions is amination has ranged from 76.2–100%, and the pre-dictive value of a negative examination has varied fromminimal9; however, the psychologic costs to the pa-

tient are unclear,10 and the monetary costs to the 46.3–98.8%. When insufficient samples are excludedfrom analysis, the sensitivity has varied from 58.3–health care system are quite high. Approximately 32%

of the total cost of breast carcinoma screening is 97.4%, the specificity has ranged from 55.5–100%, andthe predictive value of a negative examination hascaused by open surgical biopsy for benign disease.11

Less expensive, minimally invasive alternatives to ranged from 46.6–98.6%.16,17,19,20,23,24,26,32–37

Unfortunately, the definitions of an insufficientopen surgical biopsy have been developed. Fine-nee-dle (20–25-gauge) aspiration (FNA) cytology and large sample also vary across studies in which the reported

incidence of insufficient sampling ranges from 0–27%needle (14–18-gauge) core biopsy (CNB) guided byeither stereotactic radiography or sonography provide of sonographically and stereotactically obtained FNA

samples.17,19,20,23–25,32–35 Even when FNA was per-precise percutaneous sampling of clinically occultbreast lesions with minimal morbidity and without formed on surgical specimens, insufficient samples

were obtained at a rate of 7%.38 The use of other non-the disfigurement of open biopsy.12–17 The most likelycomplications (fainting, infection, and hematoma) are stereotactic localization techniques for needle place-

ment reportedly increases the number of insufficientrare, even with the use of large needles.18–20 The doseof radiation with stereotactic guidance (estimated to FNA samples.39–43 The best results have been reported

from a single center with a single person performingbe 645 mR) is not significant to the breast as a wholeand is considered acceptable for the specifically tar- the aspirations.40 Most of the other studies report only

a small number of cases. There has been a need togeted area being sampled.21 Seeding of the needletrack with malignant cells, which has been reported study FNA at a number of diagnostic centers, all per-

forming the procedure with the same high qualityfor CNB, never has been reported for breast FNA.22,23

Compared with CNB, FNA has advantages derived technique.These problems with prior studies of breast FNAfrom the smaller size of the needle; however, its gen-

eral clinical utility has been in question because of led the National Cancer Institute to initiate a largemulticenter study comparing the sensitivity and speci-variable results. Previous research has not shown

clearly whether the variability of the results reported ficity of FNA followed by CNB with that of CNB alonein the diagnosis of patients with suspicious nonpalpa-primarily is due to limitations in the technology or due

to variations in methods between the centers per- ble breast lesions. The study, which was part of a Radi-ology Diagnostic Oncology Group (RDOG) protocol,forming the procedure. The size of the needle has var-

ied from 20–23-gauge.19,20,23–26 Techniques vary from was scheduled to report results after ú 2000 patientswere evaluated. However, during an interim analysisthree passes of the needle and a coaxial approach to

two to four punctures without the coaxial tech- of the data, the Data Safety and Monitoring Board forthe study recommended closing the arm of the trialnique.15,17,19,23,25,27 No definitive studies specifically

have addressed the optimal number of passes needed that involved FNA. It did this because its membersbelieved that the overall insufficient sampling rate forfor FNA of nonpalpable lesions.28,29 Other factors that

vary from center to center include the experience of FNA was unacceptably high.In this article, we present information on the in-the operators, the types of lesions sampled, the consid-

erably different types of stereotactic device used, the sufficient sampling rate of FNA overall, by type ofmammographic lesion (calcification or mass), and byexperience of the interpreting cytopathologists, and

the methods of specimen preparation.13,17,19,20,23–25,29–31 degree of mammographic suspicion regarding thepossibility of malignancy. We also show the final diag-Previous research protocols also have varied and not

all studies have compared FNA results with histologic noses as determined from specimens obtained by FNAcompared with those obtained from CNB or surgery.proof from open surgical biopsy, nor has there been

long term clinical follow-up on all cases.19,25,30,31 Our results show that the insufficient sampling ratefor FNA is high for both masses and calcified lesions,Consequently, the results of FNA, although gener-

ally promising, have been variable. In studies in which and that a small number of patients whose FNA speci-mens were interpreted as benign or atypical by FNAFNA cytology for nonpalpable breast lesions has been

compared with open biopsy outcomes, the results had CNB or surgical specimens interpreted as malig-nant.have been as follows. When outcomes with insufficient

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

Rate of Insufficient Samples for FNA/Pisano et al. 681

TABLE 1Mammographic Criteria for Inclusion and Exclusion from the Protocol

Mammographic lesions eligible for entry onto protocolAny circumscribed solid mass over 1 cm in sizeAny circumscribed, solid mass under 1 cm in size that has grown since a prior mammogramArea of architectural distortion with a lucent center (probably a radial scar)Area of architectural distortion, unless in a known prior biopsy siteMicrolobulated noncystic mass, regardless of sizeSpiculated noncystic mass, regardless of sizeNoncystic mass with indistinct bordersNoncystic mass associated with suspicious calcifications, architectural distortion, focal skin thickening, or a focal solitary dilated ductClustered calcifications that are pleomorphic, linear, branching, variable in size, and/or very small and granularAsymmetry associated with a noncystic mass, suspicious calcifications, architectural distortion, a solitary dilated duct and/or focal skin thickeningMammographic lesions specifically excluded from entry onto protocolSingle dilated duct not associated with spontaneous nipple dischargeLocalized cluster of tiny round or oval uniform calcificationsCircumscribed, low density mass õ1 cm in size unless changed since prior mammogramMass containing intralesional fat density that is pathognomonic for lymph node, lipid (oil) cyst, or hamartomaLocalized focal asymmetric area of fibroglandular densitySubtle area of architectural distortion without central increased fibroglandular density in an area of prior biospyDiscrete cluster or clusters of calcifications suggestive of milk of calcium, secretory disease, or sclerosing adenosisDiffuse nonclustered calcifications suggestive of milk of calcium (a mammographically pathognomonic type of fibrocystic change), secretory disease, or sclerosing adenosisMultiple noncalcified, well defined massAny lesion found to be a cyst at ultrasound

MATERIALS AND METHODS from entry into the protocol are listed in Table 1. Ifa patient had more than one lesion, only the mostPatient Recruitment and Randomization

Eighteen institutions enrolled patients in 1 of 2 arms suspicious lesion was biopsied as part of the study. Allpreenrollment mammograms were reviewed for eligi-of the protocol. Needle biopsies using stereotactic

guidance only were performed at Arm A institutions bility by one of two radiologists (E.D.P. or L.L.F.). Apatient also was excluded from entry into the study if(University of Pennsylvania; Henry Ford Hospital;

Bellevue Hospital; New York University Hospital; Case her lesion was located in a prior lumpectomy or radia-tion therapy site; if she had a known bleeding disorderWestern Reserve University; New England Deaconess

Hospital; Yale University; Providence Imaging of An- or was receiving anticoagulant therapy; or if she waspregnant, allergic to local anesthesia, had breast im-chorage, AK; and Madigan Army Hospital). Needle bi-

opsies using either stereotactic or sonographic guid- plants, or had a psychiatric or neurologic conditionthat would limit her ability to cooperate during theance were performed at Arm B institutions (Baylor-

Komen Breast Center of Dallas.; Breast Diagnostic biopsy and/or provide acceptable informed consent.After the patient agreed to participate in the studyClinic of Waukesha, WI; Emory University; Johns Hop-

kins University; Radiology Imaging Associates of Boul- and informed consent was obtained, a randomizationassignment was given by the American College of Ra-der, CO; Thomas Jefferson University Hospital;

University of Maryland; University of North Carolina- diology Statistical Center in Philadelphia through atelephone call or by electronic mail. For institutionsChapel Hill; and University of Virginia). All participat-

ing centers were required to confirm that at least 50 assigned to Group A, patients were randomized andassigned either to stereotactically guided FNA (SFNA)FNAs of the breast had been performed prior to ac-

crual of patients onto the protocol by the participating and stereotactically guided CNB (SCNB) or SCNB only.For Group B institutions, patients were randomizedradiologists to minimize the possibility of poor perfor-

mance of this procedure because of inadequate experi- either to ultrasound-guided FNA (USFNA) and ultra-sound-guided CNB (USCNB), or USCNB only, or SFNAence. There were no limitations on the number of radi-

ologists who could participate at each center or the and SCNB, or SCNB only (Fig. 1). Crossover from theassigned imaging guidance system to the nonassignedlength of time of their prior experience with FNA.

Consecutive women whose nonpalpable lesions system was allowed if the lesion was inaccessible tostereotactic biopsy or not visible by sonography. Thus,were detected through mammography within 3

months of entry into the study were eligible at the if a patient was assigned to SFNA followed by SCNB,but the lesion lay in a location so close to the chestparticipating institutions. The types of mammo-

graphic lesions that were eligible for and excluded wall that it was inaccessible to stereotactic biopsy, the

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

682 CANCER February 15, 1998 / Volume 82 / Number 4

patient would undergo USFNA followed by USCNB.Similarly, if a lesion was not visible by ultrasound be-cause it was a cluster of calcifications and the patientwas assigned to USCNB only, the patient would un-dergo SCNB only.

FNA ProcedureThe RDOG5 FNA protocol required a minimum of fivepasses into the lesion with imaging guidance with a22-, 23-, or 25-gauge needle. Coaxial guidance systemswere permitted. Cytopathologists were permitted toperform the aspiration itself at the discretion of theradiologist and cytopathologist. Both standard FNAwith suction and without suction were permitted.

For masses, samples were obtained from the mid-dle of the lesion and along radii extending toward the

FIGURE 1. Radiologic Diagnostic Oncology Group 5 study scheme.12, 3, 6, and 9 o’clock positions. For calcifications,SCNB: stereotactically guided core needle biopsy; SFNA: stereotacticallyspecific calcifications were targeted. For SFNA, stereo-guided fine-needle aspiration biopsy; USFNA: ultrasonographically guidedradiographs were obtained after the placement of bothfine-needle aspiration biopsy; USCNB: ultrasonographically guided corethe first and the last needles. For USFNA, real-timeneedle biopsy. Brackets indicate crossover allowed in both directions.sonographic observation of the needle was performed

by the radiologist in two planes to assess the accuracyof the desired position.

For sites using a coaxial technique, the first three (or balanced salt solution) rinse of the needles andthose were stained with the Papanicolaou method. Atpasses with the fine needle were made in three differ-

ent planes along the path of the guide needle and a each site, the local cytopathologist interpreted all cyto-logic specimens for routine clinical use according tosecond pass of the guide needle was made into a differ-

ent part of the lesion for the localization for two addi- the individual institution’s protocol.All cytologic specimens and cytology reports weretional passes. Greater than five passes were made at

the discretion and direction of the radiologist or the reviewed by the RDOG5 study’s central cytopatholo-gists (N.S. or W.J.F.). To standardize FNA data re-on-site cytopathologist, if one was available.

At some centers, the aspirated material was porting across all institutions, the local cytopatholo-gist’s interpretation was coded by the central cytopa-smeared immediately by the radiologist, cytopatholo-

gist, or an assistant. Both air-dried and wet-fixed slides thologist. The central reader read and coded the localreport only after rendering his or her own interpreta-were made of each pass. At other centers, the primary

specimen was placed in a solution of 50% Ringer’s tion of the specimen. A specimen was considered in-sufficient for interpretation if the aspirate containedlactate/50% alcohol, Cytolite (or a balanced salt solu-

tion), and sent to the cytology laboratory for Millipore- fewer than six epithelial cell groups on all slides pro-vided for central review. The central reader coded bothfilter or cytospin preparation. At centers at which spec-

imens were smeared immediately, the sample re- the local and central interpretation using the followingscale: insufficient, benign, atypical, probably malig-maining in the needle after smearing also was placed

into one of these solutions. The availability of on-site nant, and malignant. If the central reader could notdetermine the local interpretation from the cytologycytopathologists for immediate handling and interpre-

tation of the specimen varied by institution. report, it was coded as undetermined. In addition, thecentral reader assessed the local site’s compliance withThe slides and fixative rinse were sent to the cytol-

ogy laboratory as soon as possible after the procedure. the RDOG 5 FNA protocol. If there was disagreementbetween the local and central cytopathologist, a read-All alcohol- or spray-fixed bloody smears were placed

in Carnoy’s solution (acetic acid 1:6 and 70% ethanol) ing by a second central cytopathologist was obtainedand the majority opinion was used for the purpose offor approximately 5 minutes before staining to lyse the

red blood cells. The air-dried slides were stained with the analysis.Diff-QuikR or other Romanowsky methods. The wet-fixed slides were stained with the Papanicolaou Methods of Analysis

Two major analyses were performed. The first involvedmethod or by hematoxylin and eosin. Millipore filtersor cytospin smears were prepared from the Cytolite determining the insufficient sampling rate overall and

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

Rate of Insufficient Samples for FNA/Pisano et al. 683

then the rates according to mammographic character- of lesion size and level of mammographic suspicion,istics of the lesions. The second analysis involved com- a logistic regression model including those variablesparing the final pathologic diagnoses obtained by FNA was employed.with those obtained by CNB or surgery. Samples ob- The odds ratio estimates in this report are basedtained using either sonographic or stereotactic guid- on the logistic regression models employed. Normalance were grouped together for analysis. approximation theory was used to construct 95% con-

FNA samples were considered insufficient if the fidence intervals (95% CI) for the odds ratios as wellcentral cytopathologist rated them as insufficient as to construct corresponding P values. All CIs for ratesbased on review of the slides. Rates of insufficient sam- of insufficient samples are based on the normal ap-ples were calculated for the entire population of pa- proximation except for the Clopper-Pearson46 inter-tients as well as for the two main lesion subgroups: vals, which are based on the binomial distribution.masses and calcified lesions. Lesions were said to be The latter were used when either the size of the samplecalcified if they had one of the following two character- or the rate of insufficient samples was small.istics: 1) calcifications without any mass or 2) a focalasymmetric density associated exclusively with suspi-cious calcifications. All other lesions were said to be RESULTSmass lesions. RDOG5 Patient Demographics and Compliance with the

For both types of lesions, comparisons of the rate RDOG 5 FNA Protocolof insufficient samples were made according to the As of November 1, 1996, 2349 women were enrolledenrolling radiologist’s level of suspicion for malig- in the trial at 20 participating institutions. (Two insti-nancy based on the entry mammogram (low vs. high). tutions entered the trial after the FNA arms of theLesions classified as low suspicion were considered trial were closed.) Of the women enrolled, 76.3% wereprobably benign or indeterminate by the enrolling ra- white, 16.9% were African-American, 4.3% were His-diologist. Lesions considered probably malignant or panic, 1.3% were Asian, and the remaining 1.3% weremalignant by the enrolling radiologist were classified other races or were not classified. Enrollees comprisedas high suspicion lesions.

1295 women (55.1%) age 35–54 years, 868 womenThe histopathologic type was determined by open

(37.0%) age 55–74 years, and 136 women (5.8%) agebiopsy (if one had been performed) or by CNB when ¢ 75 years. The remaining 2.1% of participants wereopen biopsy had not been performed. Lesions were

age õ 35 years. As per the protocol, patients whoseconsidered atypical if biopsy showed atypical lobular

mammographic lesions had been judged probably be-hyperplasia, atypical ductal hyperplasia, or lobular

nign or indeterminate and whose needle biopsy resultscarcinoma in situ. Lesions were considered malignantwere definitively benign were assigned to mammo-if biopsy showed ductal carcinoma in situ or invasivegraphic follow-up rather than open biopsy; thus, ap-carcinoma. All other lesions were considered benign.proximately 80% of the patients went to follow-upChi-square tests were used to compare rates ofrather than open biopsy after their needle biopsy.insufficient samples for the two lesion types and the

There were 419 patients from 18 institutions whotwo levels of suspicion. A logistic regression modelunderwent FNA, 42 of whom (10.0%) had outcomesthen was fit for mass lesions, relating the probabilityconsidered ineligible for analysis because of noncom-of insufficient sampling for these lesions to the size ofpliance with the protocol. Specifically, 22 lesions werethe lesion and the level of mammographic suspicionsampled using the wrong needle gauge (19 used a 20-for malignancy. To investigate the possible nonlineargauge needle, 1 used a 21-gauge needle, and 2 had nodependence of the log (odds) ratio of insufficiency onreported needle gauge), and in 20 patients,õ 5 passesthe size of the lesion, a nonparametric logistic regres-were performed without the input of an on-site cyto-sion, which models the logarithm of the odds ratio aspathologist. After excluding these cases, we examineda cubic smoothing spline function of lesion size,44 wasthe rate of insufficient samples for the remaining 377employed in this analysis. For calcified lesions a logis-women, among whom there were 236 mass lesionstic regression model45 was employed that related theand 141 calcified lesions.probability of obtaining an insufficient sample to the

Of the patients who are included in this analysis,number of calcifications associated with the lesion (°82% were white, 12.5% were African-American, 4%15 vs. ú 15 calcifications) and to the level of mammo-were Hispanic, 1% were Asian and 0.5% were othergraphic suspicion of the lesion.races. The study population was 53% women betweenThe rates of insufficient samples for the two guid-the ages of 35–54 years, 37% women age 55–74 years,ance systems (stereotactic and sonographic) also were

compared. To adjust for possible confounding effects and 8% women age ú 75 years.

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

684 CANCER February 15, 1998 / Volume 82 / Number 4

TABLE 2Fine-Needle Aspiration Cytology Results by Lesion Type

FNA resultsNo. (% of lesion type)

Insufficient Low grade/Lesion type sample Benign Atypical DCIS Malignant Total

Mass 58 (26.6)a 121 (55.5) 8 (3.7) 11 (5.1) 20 (9.2) 218Calcifications 65 (46.1)a 60 (42.6) 4 (2.8) 4 (2.8) 8 (5.7) 141Mass and

calcifications 5 (27.8) 5 (27.8) 0 1 (5.6) 7 (38.9) 18Total 128 (34.0) 186 (49.3) 12 (3.2) 16 (4.2) 35 (9.3) 377

FNA: fine-needle aspiration; DCIS: ductal carcinoma in situ.a P õ 0.001.

Overall Rate of Insufficient Samples FNA samples, whereas 18 calcified lesions of high sus-picion produced only 3 (16.7%; 95% CI, 3.6–41.4%)Overall, FNA yielded 128 insufficient samples for the

377 eligible patients (33.95%; 95% CI, 29.2–38.7%). The insufficient samples. Therefore the odds ratio for aninsufficient sample was 5.1 times higher for low suspi-cytopathologists overreading the cases noted on re-

view that most of the insufficient samples were acellu- cion calcified lesions than for high suspicion calcifiedlesions (95% CI, 1.4–18.4; P Å 0.013). Furthermore, forlar with only rare cases having too few cell clusters to

be considered sufficient. Two institutions contributing calcified lesions of low suspicion, the odds ratio ofobtaining an insufficient sample was 2.7 times highera total of 33 and 13 cases, respectively, to the trial

had rates of insufficient samples of 3.3% and 7.7%, for lesions with ° 15 calcifications than for calcifiedlesions with ú 15 calcifications (95% CI, 1.3–5.6; P õrespectively, but all other centers had rates of insuffi-

cient samples between 11.1–80.0%. 0.01). Forty-three (60.6%; 95% CI, 49.2–71.9%) of the71 low suspicion calcified lesions that had ° 15 calci-fications had insufficient FNA samples, whereas 19Rate of Insufficient Samples by Type of Lesion and Level

of Suspicion (36.5%; 95% CI, 23.5–49.6%) of the 52 low suspicioncalcified lesions that hadú 15 calcifications had insuf-As is seen in Table 2, the rate of insufficient samples

varied by type of lesion with calcified lesions (as de- ficient samples. For high suspicion calcified lesions,the number of calcifications was not a statistically sig-fined earlier) associated with a significantly higher rate

of insufficient sampling than masses (P õ 0.001). Of nificant factor, but there were very few high suspicioncalcified lesions entered into the study.the 236 masses, 29.2% (95% CI, 23.8–34.7%) yielded

insufficient samples, compared with 46.1% (95% CI,37.9–54.3%) of the calcified lesions. Masses with calci- Rate of Insufficient Samples by Histopathologic

Diagnosisfications had an insufficient sampling rate similar tomasses without calcifications (27.8% vs. 26.6%) and We also compared the rate of FNA insufficient samples

with the histology of the lesion being sampled as deter-therefore were grouped together for the remainder ofthis study. mined either by CNB or open surgical biopsy. Table 3

shows that for the 336 lesions for which histologicThe radiologist’s level of suspicion of the lesionwas not a statistically significant predictor of insuffi- information was available, insufficient samples oc-

curred in 34.2% (95% CI, 26.8–41.5%) of the benigncient samples for mass lesions, but was a predictorfor calcified lesions. With 54 of the 184 low suspicion masses, but in significantly fewer of the malignant

masses (12.5%; Clopper-Pearson 95% CI, 4.2–26.8%; Pmasses (29.3%) having insufficient FNA samples, aswell as 9 of the 52 high suspicion masses (17.3%), the Å 0.007). Likewise, samples were insufficient in sig-

nificantly more of the benign clusters of calcificationsodds ratio was 2.0 (95% CI, 0.9–4.4; P .09). The size ofthe mass lesion also was not found to be a statistically (55.9%; 95% CI, 46.2–65.5%) than in the malignant

clusters (33.3%; Clopper-Pearson 95% CI, 14.6–57.0%;significant predictor of insufficient FNA samples (P Å0.41). P Å 0.06).

If atypical histologies were combined with malig-Of the 123 calcified lesions classified as low suspi-cion, 62 (50.4%; 95% CI, 41.6–59.2%) had insufficient nant ones and compared with benign lesions, insuffi-

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

Rate of Insufficient Samples for FNA/Pisano et al. 685

TABLE 4TABLE 3FNA Sample Sufficiency According to Histopathology and Lesion Type FNA Sample Sufficiency by Cytopathologist Status and Lesion Type

FNA resultsFNA resultsNo. (% of histologic type) No. (% of category)

On-siteLesion type cytopathologist Sufficient Insufficient TotalHistologic

Lesion type results Sufficient Insufficient TotalMass No 119 (68.8) 54 (31.2) 173

Yes 53 (85.5) 9 (14.5) 62Mass Benign 106 (65.8) 55 (34.2)a 161ALH/LCIS/ Calcifications No 55 (50.0) 55 (50.0) 110

Yes 20 (74.1) 7 (25.9) 27ADH 2 (100.0) 0 2DCIS/

invasive FNA: fine-needle aspiration.

carcinoma 35 (87.5) 5 (12.5)a 40Total 143 (70.4) 60 (29.6) 203

Calcifications Benign 45 (41.1) 57 (55.9)b 102ALH/LCIS/ an on-site cytopathologist significantly improved the

ADH 6 (85.7) 1 (14.3) 7 frequency of sufficient FNA sampling for both massDCIS/

lesions and calcifications (odds ratio for masses, 2.7,invasive95% CI, 1.2–5.8; P Å 0.028 vs. odds ratio for calcifica-carcinoma 14 (66.7) 7 (33.3)b 21

Total 65 (50.0) 65 (50.0) 133 tions, 2.9, 95% CI, 1.1–7.3; P Å 0.013). However, thisTotal 208 (62.5) 125 (37.5) 336 conclusion cannot necessarily be drawn. Because each

institution either had an on-site cytopathologist or didFNA: fine-needle aspiration; ALH: atypical lobular hyperplasia; LCIS: lobular carcinoma in situ; ADH:

not, the results cannot be separated from other institu-atypical ductal hyperplasia; DCIS: ductal carcinoma in situ.tional effects.a P Å 0.007.

b P Å 0.06.

Rate of Insufficient Samples by Guidance SystemThere was extensive crossover of patients from theassigned guidance system, mostly from ultrasound to

cient samples occurred in 5 of 42 masses (11.9%; Clop- stereotactic guidance. Crossover was allowed at insti-per-Pearson 95% CI, 4.0–25.6%) and 8 of 28 calcified tutions in which both guidance systems were availablelesions (28.6%; Clopper-Pearson 95% CI, 13.2–48.7%). (Group B) for lesions not visible or accessible by theThe P value when these rates were compared with the assigned method. Group B institutions accounted forinsufficient rate for benign masses was 0.005 and was 246 of 377 FNAs performed on the protocol, including0.01 for benign calcifications. 92 patients with calcifications and 154 with masses.

Of the 116 low suspicion lesions that yielded insuf- Only 1 (1.9%) of the 52 calcified lesions assigned toficient FNA specimens, 99 (85.3%) showed benign his- USFNA (plus CNB) actually was biopsied with ultra-tology by CNB, and subsequently were submitted to sound guidance. None of the 40 subjects with calcifi-follow-up mammography only. Of the 17 low suspi- cations who were assigned to stereotactic guidancecion lesions with insufficient FNA specimens that were crossed over to sonographic guidance. Of the 79 pa-submitted to open surgical biopsy, 7 proved to be ma- tients with masses assigned to SFNA plus CNB, 27lignant (41%). Of the 12 high suspicion lesions with (34.2%) crossed over to sonographic guidance,insufficient FNA samples, 8 (67%) subsequently were whereas 32 of the 75 patients with masses (42.7%) as-been submitted to open surgical biopsy. Of those, only signed to USFNA plus CNB were crossed over to ste-1 lesion(12.5%) proved to be malignant. Even if all reotactically guided biopsy.cases in follow-up ultimately proved to be benign, the The overall insufficient sampling rates were 122rate of malignant lesions with insufficient FNA sam- of 306 (39.9%) for stereotactic guidance, and 6 of 71ples would be 7 of 116 (6.0%) for low suspicion findings (8.45%) for ultrasound guidance. For mass lesions, in-and 1 of 12 (8.3%) for high suspicion findings. sufficient samples were obtained from 57 of 166 le-

sions (34.3%) aspirated using stereotactic guidanceand 6 of 70 lesions (8.6%) aspirated using sonographicRate of Insufficient Samples and Cytopathologist’s

Availability during the Procedure guidance. For calcified lesions, 65 of 140 lesions(46.4%) aspirated with stereotactic guidance gave in-In Table 4, the rates of insufficient samples for masses

and calcified lesions are given according to whether a sufficient samples. Sonographic guidance for the onecalcified lesion sampled this way yielded a sufficientcytopathologist was present during the FNA proce-

dure. It appears from these data that the presence of sample.

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

686 CANCER February 15, 1998 / Volume 82 / Number 4

There was a subset of cases that yielded low rates samples. Consequently, this factor cannot be used reli-ably to determine which patients should proceed toof insufficient samples. Patients who might have had

cysts were required to undergo prestudy sonography. CNB or open surgical biopsy when an insufficient FNAoccurs.For high suspicion masses that were visible on

prestudy ultrasound, the rate of insufficient samples The rate of insufficient samples was very high atthe majority of the centers participating in this trial.was 1 of 17 (5.9%) for women who underwent USFNA

and for 0 of 7 (0%) for women who underwent SFNA. The centers that participated in this study includedacademic and private institutions with board-certifiedFor women with low suspicion masses that were visi-

ble on prestudy ultrasound, the rate of insufficient radiologists and pathologists with prior experienceperforming this procedure. The high rate of insuffi-samples was 5 of 52 (9.6%) for women who underwent

USFNA. In contrast, the rate of insufficient samples cient samples likely reflects how this procedure per-forms in the community at large. Although FNA forwas 22 of 51 (43.1%) for women with low suspicion

masses visible on prestudy ultrasound who underwent palpable lesions may be cost-effective with rates ofinsufficient samples of ° 10%,49,50 the additional costSFNA.of imaging guidance required for nonpalpable lesionsmakes the procedure less cost-effective in general.DISCUSSION

One goal of RDOG5 was to determine whether FNA of Trends observed in the current study may havebearing on the design for any future studies of FNA ofthe breast can replace open surgical biopsy for nonpal-

pable breast lesions. If percutaneous biopsy had not nonpalpable breast lesions. First, when a cytopatholo-gist was on site, the rate of insufficient sampling wasbeen developed, the women entered in this study

probably would have undergone open surgical biopsy. cut in half. Because this factor was not randomized inthis study, it cannot be determined whether theInsufficient samples yielded no information that

would allow for follow-up rather than surgery. The change was caused by the cytopathologist’s presenceor other undefined institutional effects. If the successhigh rate of insufficient samples for FNA in this setting,

as was present in this multicenter trial, makes its use of FNA were to depend on the former factor, the over-all utility of the procedure would require review inimpractical. A rate of insufficient samples of 10%

might make the procedure clinically useful.35,49,50 light of its increased cost and reduced availability.Second, the results of the current study show aTo our knowledge, no prior studies of this technol-

ogy involved the large number of centers and opera- tendency toward lower insufficiency rates with sono-graphic than with stereotactic guidance. Becausetors used in this study. Central overreading of cytopa-

thology specimens by experts allowed for standardiza- crossover between guidance systems was allowed inthis study, the lesions finally biopsied with US guid-tion of interpretation for this study that was not

present in other published trials. The large number ance represent a highly selected subgroup, creatingtoo much bias for formal comparison of guidanceof cases entered into the trial using a standardized

protocol for aspirating, processing, and interpreting techniques. It may be the case that FNA limited to justpatients who are easily amenable to sonographicallyspecimens also is unique to this study.

The FNAs conducted in this study yielded samples guided biopsy has clinical utility without yielding largenumbers of insufficient samples.sufficient for cytopathologic analysis most reliably for

patients whose mammographically detected lesions Third, some cytopathologists define insufficientsamples as those containing fewer than four to sixwere masses rather than calcifications, regardless of

the extent of calcifications associated with the mass, groups of cells. Although this definition differs fromthat used in the current study, in which an insufficientthe radiologist’s level of suspicion of the mass, or the

size of the mass. Furthermore, FNA samples were sample was one that included six groups of cells orfewer, this change in definition would not havemore often sufficient in cases in which the lesion was

found to be malignant or histopathologically atypical changed our results significantly. When samples wereinsufficient in this study, they usually were acellularrather than benign, both for masses and calcifications.

Calcified lesions of low suspicion with ° 15 calcifica- or had only one or two groups of cells.Finally, no special training for FNA or specimentions had significantly greater odds for yielding an in-

sufficient FNA sample. Calcified lesions more often preparation was provided to the clinicians as part ofthis study. Additional training might reduce the rateyielded insufficient samples than did masses.

Even if all cases in follow-up ultimately proved to of insufficient samples.47

Imaging-guided percutaneous FNA for nonpalpa-be benign, the suspiciousness of the lesion by mam-mography did not significantly alter the probability ble breast lesions yielded too many insufficient sam-

ples to be deemed clinically practical in this multiinsti-of malignancy for patients who have insufficient FNA

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

Rate of Insufficient Samples for FNA/Pisano et al. 687

reotactic localization and fine-needle aspiration. Radiologytutional study. This likely reflects how the procedure1989;170:427–33.performs in most centers within the larger community.

21. Harter LP, Curtis JS, Ponto G, Craig PH. Malignant seedingof the needle track during stereotactic core needle breastbiopsy. Radiology 1992;185:713–4.REFERENCES 22. Smith EH. The hazards of fine-needle aspiration biopsy. US

1. Feig SA. Decreased breast cancer mortality through mam-Med Bio 1984;10:629–34.

mographic screening: results of clinical trials. Radiology23. Fajardo L, Davis J, Wiens J, Trego D. Mammography guided1988;167:659–65.

stereotactic fine-needle aspiration cytology of nonpalpable2. Shapiro S. Evidence of screening for breast cancer from abreast lesions. AJR Am J Roentgenol 1990;155:977–81.randomized trial. Cancer 1977;39:2772–82.

24. Dowlatshahi K, Yaremko L, Kluskens L, Jokich P. Nonpalpa-3. Shapiro S, Venet W, Strax P, Venet L, Roeser R. Ten-to-four-ble breast lesions: findings of stereotactic needle core biopsyteen-year effect of screening on breast cancer mortality. Jand fine-needle aspiration cytology. Radiology 1991;181:Natl Cancer Inst 1982;69:349–55.745–50.4. Schwartz G, Feig S, Patchefsky A. Significance and staging of

25. Ciatto S, Rosselli del Turco M, Bravetti P. Nonpalpable breastnonpalpable carcinomas of the breast. Surg Gynecol Obstetlesions: stereotactic fine-needle aspiration cytology. Radiol-1988;166:6–10.ogy 1989;173:57–9.5. Skinner M, Swain M, Simmon R, McCarty K, Sullivan D,

26.Iglehart J. Nonpalpable breast lesions at biopsy. Ann Surg Bibbo M, Scheiber M, Cajulis R, Keebler C, Wied G, Dowlats-1988;208:203–8. hahi K. Stereotactic fine-needle aspiration. Cytology of clini-

6. Landercasper J, Gunderson S, Gunderson A, Cogbill T, Trav- cally occult malignant and premalignant breast lesions. Actaelli R, Strutt P. Needle localization and biopsy of nonpalpa- Cytol 1988;32(2):193–201.ble lesions of the breast. Surg Gynecol Obstet 1982;164:399– 27. Mitnick JS, Vazquez MR, Roses DF, Harris MN, Granutsos403. R, Waisman J. Stereotactic localization for fine-needle aspi-

7. Wright C. Breast cancer screening. Surgery 1986;100(4): ration breast biopsy. Initial experience with 300 patients.594–8. Arch Surg 1991;126:1137–40.

8. U.S. Preventive Services Task Force. Guide to clinical pre- 28. Pennes DR, Naylor B, Rebner M. Fine-needle aspiration bi-ventive services: an assessment of the effectiveness of 169 opsy of the breast. Influence of the number of passes andinterventions. Baltimore: Williams and Wilkins, 1989:39–62. the sample size on the diagnostic yield. Acta Cytol 1990;

9. Sickles E. Periodic mammographic follow-up of probably 34(5):673–6.benign lesions: results in 3184 consecutive cases. Radiology 29. Jackson VP, Reynolds HE. Stereotactic needle-core biopsy1991;179(2):463–8. and fine-needle aspiration cytologic evaluation of nonpalpa-

10. Helvie MA, Ikeda DM, Adler DD. Localization and needle ble breast lesions. Radiology 1991;181:633–4.aspiration of breast lesions: complications in 370 cases. AJR

30. Dronkers D. Stereotactic core biopsy of breast lesions. Radi-Am J Roentgenol 1991;157:711–4.ology 1992;183:631–4.11. Cyrlak D. Induced costs of low-cost screening mammogra-

31. Elliott RL, Haynes AE, Bolin JA, Boagni EM, Head JF. Stereo-phy. Radiology 1988:661–3.tactic needle localization and biopsy of occult breast lesions:12. Schmidt R, Morrow M, Bibbo M, Cox S. Benefits of stereotac-first year’s experience. Am Surg 1992;58:126–31.tic aspiration cytology. Admin Radiol 1992;9:35–42.

32. Svane G, Silfversward C. Stereotaxic needle biopsy of non-13. Parker S, Lovin J, Jobe W, Luethke J, Hopper K, Yakes W, etpalpable breast lesions. Acta Radiol 1983;24:283–8.al. Stereotactic breast biopsy with a biopsy gun. Radiology

33. Dent DM, Kirkpatrick AE, McGoogan E, Chetty U, Anderson1990;176:741–7.14. Parker S, Lovin J, Jobe W, Burke B, Hopper K, Yakes W, et TJ. Stereotaxic localization and aspiration cytology of impal-

al. Nonpalpable breast lesions: stereotactic automated large- pable breast lesions. Clin Radiol 1989;40:380–2.core biopsies. Radiology 1991;180:403–7. 34. Layfield LJ, Parkinson B, Wong J, Giuliano AE, Bassett LW.

15. Sterrett G, Oliver D, Trayne J, Ingram D, Sherner H. Stereo- Mammographically guided fine-needle aspiration biopsy oftactic fine-needle aspiration biopsy (SFNB) of breast: pre- nonpalpable breast lesions. Can it replace open biopsy?liminary results in Perth with the TRC Mammotest Machine. Cancer 1991;68:2007–11.Cytological aspects. Pathology 1991;23:302–10. 35. Sneige N, Fornage BD, Saleh G. Ultrasound-guided fine-nee-

16. Dowlatshahi K, Jokich PM, Schmidt R, Bibbo M, Dawson PJ. dle aspiration of nonpalpable breast lesions. Cytologic andCytologic diagnosis of occult breast lesions using stereotac- histologic findings. Am J Clin Pathol 1994;102(1):98–101.tic needle aspiration. A preliminary report. Arch Surg 1987; 36. Fornage BD, Faroux MJ, Simatos A. Breast masses: US-122:1343–6.

guided fine-needle aspiration biopsy. Radiology 1987;162:17. Azavedo E, Svane G, Auer G. Stereotactic fine-needle biopsy

409–11.in 2594 mammographically detected nonpalpable lesions.

37. Gordon PB, Goldenberg SL, Chan NHL. Solid breast lesions:Lancet 1989;XX: 1033–5.diagnosis with US-guided fine-needle aspiration biopsy. Ra-18. Evans WP, Cade SH. Needle localization and fine-needlediology 1993;189:573–80.aspiration biopsy of nonpalpable breast lesions with use of

38. Teixidor H, Wojtasek D, Reiches E, Santos-Buch C, Minickstandard and stereotactic equipment. Radiology 1989;173:R. Fine-needle aspiration breast biopsy specimens: correla-53–6.tion of histologic and cytologic findings. Radiology 1992;19. Lofgren M, Anderson I, Lindholm K. Stereotactic fine-needle184:55–8.aspiration for cytologic diagnosis of nonpalpable breast le-

39. Hahn L, Ducatman B, Wang H, Fein V, McIntire J. Nonpalpa-sions. AJR Am J Roentgenol 1990;154:1191–5.ble breast lesions: evaluation by means of fine-needle aspi-20. Dowlatshahi K, Gent H, Schmidt R, Jokich P, Bibbo M,

Sprenger E. Nonpalpable breast tumor: diagnosis with ste- ration cytology. Radiology 1989;171:373–6.

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer

688 CANCER February 15, 1998 / Volume 82 / Number 4

40. 45. Hosmer HW, Lemeshow S. Applied logistic regression. NewMasood S, Frykberg ER, McLellan GL, Scalapino MC, Mit-York: John Wiley and Sons, Inc., 1989.chum DG, Bullard JB. Prospective evaluation of radiographi-

46. Korn EL. Sample size tables for bounding small proportions.cally directed fine-needle aspiration biopsy of nonpalpableBiometrics 1986;42:213–6.breast lesions. Cancer 1990;66:1480–7.

47. Hayes MK, DeBruhl ND, Hirschowitz S, Kimme-Smith C,41. Arishita GL, Cruz BK, Harding CT, Arbutina DR. Mammo-Bassett LW. Mammographically guided fine-needle aspira-gram-directed fine-needle aspiration of nonpalpable breasttion cytology of the breast: reducing the rate of insufficientlesions. J Surg Oncol 1991;48:153–7.specimens. AJR Am J Roentgenol 1996;167:381–4.42. Lofgren M, Andersson I, Bondeson L, Lindholm K. X-ray

48. Klein TS. Adequacy and aspirates from the breast: a philo-guided fine-needle aspiration for the cytologic diagnosis ofsophical approach. Diagn Cytopathol 1995;13:470–2.nonpalpable breast lesions. Cancer 1988;61:1032–7.

49. Silverman JF, Lannin DR, O’Brien K, Norris HT. The triage43. Helvie MA, Baker DE, Adler DD, Andersson I, Naylor B, role of fine needle aspiration biopsy of palpable breast

Buckwalter KA. Radiographically guided fine-needle aspira- masses. Diagnostic accuracy and cost-effectiveness. Acta Cy-tion of nonpalpable breast lesions. Radiology 1990;174:657– tol 1987;31(6):731–6.61. 50. Lannin DR, Silverman JF, Pories WJ, Walker C. Cost-effec-

44. Hastie T, Tibshirani R. Generalized additive models: some tiveness of fine needle biopsy of the breast. Ann Surg 1986;applications. J Am Stat Assoc 1987;82:371–86. 203(5):474–80.

/ 7bb2$$0906 01-13-98 15:34:28 canxa W: Cancer