FlorenceSydney Breast Biopsy Study: sensitivity of ultrasound-guided versus
freehand ne needle biopsy of palpable breast cancer N. Houssami 1,2 , S. Ciatto 1 , D. Ambrogetti 1 , S. Catarzi 1 , G. Risso 1 , R. Bonardi 1 , and L. Irwig 2 1 Centro per lo Studio e la Prevenzione Oncologica, Florence, Italy; 2 Screening and Test Evaluation Program, School of Public Health, University of Sydney, NSW, Australia Key words: breast cancer, ne needle aspiration biopsy (FNAB), sensitivity, ultrasound-guidance Summary Background. Fine needle aspiration biopsy (FNAB) is widely used in the diagnosis of breast cancer. It is unknown whether, for palpable cancers, ultrasound-guided FNAB is more accurate than freehand FNAB, and practice varies between physicians, services and countries. Methods. From consecutive women attending a major cancer centre in Florence, we prospectively recruited subjects who had a denitely palpable lump which was solid on ultrasound and suspicious of malignancy (n 102). All subjects were investigated using both ultrasound-guided and freehand FNAB (one aspirate with each method). Radiologists skilled in both sampling techniques performed all clinical examinations and aspirations, and for each subject the same radiologist obtained both FNAB samples. Sequence of aspiration method was randomised. Cytological interpretation was blinded to method of sampling. Comparative sensitivity (and insuciency) for FNAB using the two methods was calculated in all cancers (n 97). Results. Ultrasound-guided FNAB resulted in 13.6% (522%) less insucient aspirates than freehand FNAB (v 2 7.58; p 0.006). When insucient aspirates are included and considered as negative, ultrasound-guided FNAB has a 14.6% (5.823%) or a 16.5% (7.625.4%) signicantly better sensitivity than freehand FNAB (for cytology 35 positive or cytology 45 positive respectively). When insucient aspirates are excluded from the analysis, ultrasound-guided FNAB has a 1.4% ()1.2 to 3.9%) or a 2.6% ()2.5 to 7.8%) higher sensitivity than freehand FNAB (for cytology 35 positive or cytology 45 positive respectively) but this dierence in the sensitivity of the two methods is not statistically signicant. Conclusion. Our data suggest that ultrasound-guided FNAB has better sensitivity than freehand FNAB in pal- pable breast cancer, which is predominantly an eect of a signicant reduction in insucient aspirates, but in part an eect of upgrading cytological classication of cancers. Introduction Fine needle aspiration biopsy (FNAB) has been used in breast diagnosis for several decades and shown to be an accurate test [14] with relatively low cost and no sig- nicant complications. Although the use of core biopsy has progressively complemented or even replaced FNAB [5], especially in the screen-detected setting, FNAB continues to be widely used in pre-operative biopsy of symptomatic breast tumours. The use of ultrasound guidance to perform FNAB is essential for non-palpable lesions, and is often used for palpable le- sions in dedicated breast centres. However, to date, there is no evidence that ultrasound-guided FNAB of masses that are denitely palpable results in an incre- mental gain in accuracy relative to freehand sampling. There is no consensus amongst breast experts with re- gards to this issue, although imaging experts may advocate ultrasound-guidance irrespective of lesion palpability. Variations in practice exist in dierent countries and in settings where breast FNAB is per- formed independently of imaging or outside of dedi- cated breast centres. Variations in FNAB practice are also partly related to whose role it is to perform FNAB of palpable masses (the pathologist, the radiologist, the surgeon, or the breast clinician) and are not based on evidence. Adding to this controversy is the potential nancial incentive in some countries, according to whether ne needle biopsy is performed with or without ultrasound-guidance depending on tumour palpability. We hypothesised that real-time sonographic visuali- sation of needle sampling of palpable tumours may confer diagnostic benet on the basis of two mecha- nisms. First, by specically targeting the lesion and conning the needle sampling to within the tumour mass it may avoid sampling from the surrounding tissue, thus preventing dilution of the specimen. Second, it may provide technical eciency in obtaining the sample, thus reducing insucient aspirates. We conducted a study to establish the comparative sensitivity of FNAB per- Breast Cancer Research and Treatment (2005) 89: 5559 Springer 2005 formed freehand with that of ultrasound-guided FNAB in women with palpable suspicious breast tumours. Methods The study was conducted prospectively from March 2001 to December 2003. Subjects were women with clinically palpable breast lumps attending a major Ital- ian cancer centre on the basis of self-referral, clinical referral, or screen-detected ndings. From consecutive women we included subjects who had palpable lumps which were: denitely palpable, shown to be solid on ultrasound, and warranting FNAB on the basis of clinical or imaging suspicion. In the study centre, judi- cious use of FNAB and/or core biopsy is based on the expected likelihood of malignancy. Palpable solid mas- ses are usually investigated with FNAB if they are considered to have suspicious features, and thus to have a high likelihood of malignancy; solid masses with be- nign features that warrant investigation are generally sampled with core biopsy or FNAB. On the basis of these eligibility criteria, our subjects would therefore have a high pre-test probability of breast cancer. Eligible subjects were invited to participate in the study, and verbal informed consent was obtained by the radiologist performing the assessment. Women who had breast prostheses were not recruited into the study. Each subject underwent freehand FNAB (one aspi- rate only) as well as ultrasound-guided FNAB (one aspirate only). The order (sequence) of performing the two sampling methods was randomised using a coin toss technique by the participating radiologist, who also performed the clinical examination. Current FNAB practice in the study centre is tailored to the logistics of the clinic, and FNAB of palpable masses may be done by the radiologist freehand or with ultrasound guidance depending on availability of an ultrasound unit where the patient is being examined. All FNAB sampling was performed by one of ve radiologists with established and ongoing experience in both freehand and ultra- sound-guided FNAB (each radiologist had over 5 years experience with each method), and the same radiologist performed the two sampling methods for each case. All sampling was performed using a 21 gauge needle and with remote controlled (automatic aspirator) suction, with the sampling operator directly handling the needle, which is standard practice in the study centre. Within subjects the same sampling conditions were strictly maintained for both sampling methods. FNAB specimens were prepared and reported in line with the centres standardised reporting protocols, by one of six cytopathologists. Participating pathologists were blinded to the sampling methods, but were aware that two samples were provided for each subject. Cytology was reported using a standardised categorical scale of 15 widely used in reporting breast cytology [68]. Data collected included sequence of sampling method, lesion size based on maximum diameter on ultrasound, and cytology result for each sampling method. Final histological diagnosis was obtained in all subjects. Data analysis The sensitivity and the adequacy of FNAB performed with and without ultrasound guidance was compared using McNemars v 2 test for paired proportions with a continuity correction. The 95% condence interval for the dierence between paired proportions was calcu- lated [9]. Comparative sensitivity of FNAB using the two sampling methods was calculated using dierent thresholds for classifying results as positive or negative. We used SAS (version 8.2) for cross-tabulation of data and for logistic regression modelling. We examined whether tumour size had an eect on the dierence in sensitivity (and the likelihood of insuciency) of the two sampling methods using logistic regression and taking account of the paired data using generalised estimating equations [10]. We examined insuciency in relation to aspirator. We calculated sensitivity for each test according to the sequence of performing the sampling method. Results There were 97 malignant lesions sampled in 93 subjects (two subjects with bilateral cancer and two subjects with multicentric cancer), and nine subjects had benign pathology. The mean age in subjects was 63 years (range: 3791 years). Invasive ductal cancer (invasive ductal NOS, and mixed invasive ductal and lobular) comprised the majority of our series (81.5%). The remaining histological types were: invasive lobular (8.3%), medullary (2.1%), tubular (2.1%), mucinous (2.1%), cribriform (2.1%), and other invasive breast cancers (2.1%). The absence of ductal carcinoma in situ from this series reects our subject inclusion criteria. Mean tumour size was 18.1 mm (range 1040 mm). Because the number of subjects who had benign pathology in our series was small (n 9) we did not calculate specicity. Comparative sensitivity of freehand and ultrasound- guided FNAB using dierent thresholds for considering results as positive or negative is presented in Table 1. In the study centre, FNAB cytology reported as category 3 (atypical) would be considered as positive (in that it would be further investigated), but we have presented comparative sensitivity using dierent thresholds for considering results as positive or negative to allow for variations in clinical practice. When insucient aspi- rates are included and considered as negative, ultra- sound-guided FNAB has a 14.6% (5.823%) or a 16.5% (7.625.4%) signicantly better sensitivity than freehand FNAB (Table 1). When insucient aspirates are ex- cluded from the analysis, ultrasound-guided FNAB has a 1.4% ()1.2 to 3.9%) or a 2.6% ()2.5 to 7.8%) higher N Houssami et al. 56 sensitivity than freehand FNAB but this dierence in the sensitivity of the two methods is not statistically signicant (Table 1). Absolute sensitivity (proportion of all cancers with category 5 cytology) was 10.3% (2 18.6%) higher for ultrasound-guided FNAB than for freehand FNAB (v 2 4.5; p 0.03), but this was mainly due to dierences in insucient aspirates. Ultrasound-guided FNAB resulted in 13.6% (522%) less insucient aspirates than freehand FNAB (v 2 7.58; p 0.006) (Table 1). Logistic regression analysis showed that tumour size did not have a signicant eect on the dierence in the sensitivity of the two sampling methods ( p 0.14) and there was no evidence of interaction between method of sampling and tumour size ( p 0.9). Tumour size, however, had an eect on the dierence in the proba- bility of an insucient result between the two sampling methods ( p 0.015). There was no signicant aspirator (or radiologist) eect on the dierence in the probability of an insucient result ( p 0.09). That is, comparing the two FNAB sampling methods, the likelihood of an insucient result was not signicantly inuenced by aspirator. However, for individual FNAB methods, insuciency rates varied signicantly between aspirators (v 2 3 8.44; p 0.04) for freehand FNAB only. Consistent with randomisation of sampling se- quence, freehand FNAB was performed rst in 50.5% of subjects, and ultrasound-guided FNAB was performed rst in 49.5%of subjects. Sequence of sampling had little eect on the sensitivity or the adequacy of ultrasound- guided FNAB: sensitivity was approximately 94.0% whether ultrasound-guided FNAB was performed rst or second in sequence; insuciency rate was 4.2% for rst in sequence and 6% for second in sequence. Esti- mates for freehand FNAB diered according to se- quence of performing the sampling but the dierences were not statistically signicant. Freehand FNAB had a sensitivity of 83.3% when performed as second in sam- pling sequence and a sensitivity of 75.5% when per- formed as rst in sequence, which was a non-signicant dierence of 7.8% (95% CI: )8 to 24%) and was largely due to dierences in insuciency. Freehand FNAB had an insuciency rate of 24.5% when performed as rst method, and an insuciency rate of 12.5% when per- formed as second method, which was a non-signicant dierence of 12% (95% CI: )3 to 27%). Discussion This paper focuses on palpable breast cancer, and shows that FNAB is more sensitive when performed using ultrasound guidance. This eect is largely achieved through a signicant reduction in insucient aspirates when FNAB is performed with ultrasound guidance. This is an important nding since insuciency is a major reason for the increasing abandonment of FNAB in breast diagnosis. It is also important since freehand FNAB of palpable breast masses is widely practised in various settings. In an overview of breast FNAB sensitivity was re- ported to be 87% (range: 7898%) [11]. It may therefore be argued that the sensitivity of freehand FNAB in our study, when insucient cytology is included as a nega- tive test result in the analysis, is relatively low (although within the reported range). We emphasise that our data present comparative sensitivity for each sampling method using one aspirate (one sample) only obtained with each technique. These data cannot be taken to be indicative of the true sensitivity of FNAB (where more Table 1. Comparative sensitivity, and adequacy, of ultrasound-guided and freehand FNAB Classication of cytology result as positive (negative) Ultrasound-guided FNAB sensitivity Freehand FNAB sensitivity Dierence in sensitivity (95% CI) Insucient cytology included in analysis Categories 35 positive 93.8% 79.4% 14.6% (5.823%) (12 negative) v 2 = 8.45 ( p = 0.004) Categories 45 positive 89.7% 73.2% 16.5% (7.625.4%) (13 negative) v 2 = 10.2 ( p = 0.001) Insucient cytology excluded from analysis Categories 35 positive 98.8% 97.4% 1.4% ()1.2 to 3.9%) (2 negative) v 2 = 0 ( p = 1) Categories 45 positive 93.4% 90.8% 2.6% ()2.5 to 7.8%) (23 negative) v 2 = 0.25 ( p = 0.6) Absolute sensitivity Proportion of all cancers with category 5 cytology 66.0% 55.7% 10.3% (2.0 to 18.6%) v 2 = 4.5 ( p = 0.03) Insuciency/inadequacy rate IR = % (number) of all cancers with category 1 cytology Ultrasound-guided IR = 5% (5/97) Freehand IR = 18.6% (18/97) Dierence in IR )13.6% ()5 to )22%) v 2 = 7.58 ( p = 0.006) Ultrasound-guided versus freehand breast needle biopsy 57 than one aspirate may be taken) and cannot be com- pared to sampling using multiple aspirates which is known to improve specimen yield. It is possible that the common practice of obtaining more than one FNAB sample may eliminate any dierence in accuracy be- tween the two sampling techniques (i.e. may overcome the incremental gain in accuracy obtained with the use of ultrasound-guidance). Were this the case, this does not invalidate our ndings since ultrasound-guidance would allow a faster and more ecient FNAB proce- dure, with less samples required (and therefore less dis- comfort to the patient) and a lower number of smears to be read. The clinical question our study addresses relates to tumours that are readily palpable. Tumours that are vaguely or inconsistently palpable, or palpable within nodular tissue or surrounding tissue reaction, were not included. We feel that it would not have been appro- priate to include this category of tumours in our study since tumours that are dicult to palpate clinically should always be sampled with image guidance to en- sure that the lesion is appropriately targeted. It is possible that radiologists participating in our study may be more skilled at ultrasound-guided FNAB or may have a preference for ultrasound-guided sam- pling, and it may be argued that this has caused the dierence in insucient aspirates in our data. While we acknowledge this potential limitation, we point out that the radiologists in our study are experienced with free- hand FNAB and continue to practice both freehand and ultrasound freehand FNAB in their work setting (about 3040% of palpable masses are sampled freehand in the study setting). Our analyses of insuciency did not show a statistically signicant radiologist (aspirator) eect on the dierence in the probability of an insu- cient result (comparing the two methods). However, we did nd signicant variations in insuciency rates be- tween aspirators for freehand FNAB only. It is possible, then, that ultrasound-guidance may minimise the eect of inherent dierences in aspirator skills. That is, ultrasound-guidance may reduce operator-related vari- ations in FNAB acquisition, potentially equalising adequacy of aspirates. Future studies should explore this aspect of our ndings, using a larger sample of subjects, and more aspirators with varying FNAB experience. We did not nd a sampling sequence eect with ultrasound-guided FNAB, but for freehand FNAB sensitivity was higher (predominantly due to lower insuciency) when freehand was performed second in the sampling sequence this was not statistically sig- nicant. This unexpected nding may be due to chance. Another possibility is that when freehand FNAB sam- pling is performed following on from ultrasound-guided sampling, there is a ow-on eect from the prior aspiration. This may be a physical eect (increased cel- lular discohesion from the ultrasound-guided aspira- tion) or an information eect whereby the aspirator has retained an image of the lesion (gained from the ultra- sound-guided aspiration) which then assists the freehand sampling. It may be argued that ultrasound-guided FNAB cannot be performed in all settings, as ultrasound is not always available and/or aspirators may not be skilled in ultrasound-guided biopsy. However, we feel that clini- cal and imaging correlation is mandatory in contem- porary diagnosis of breast diseases, and ultrasound should always be accessible wherever dierential diag- nosis of breast abnormalities is being undertaken. In addition, skills in ultrasound-guided biopsy may be acquired by clinicians of varying disciplines. Our paper has presented the rst evidence on the comparative sensitivity of ultrasound-guided and free- hand FNAB in women with palpable breast cancer. We used a paired study design that allows comparison within subjects, thus maintaining patient and tumour variables constant for both sampling methods. Cyto- logical interpretation of FNAB was blinded to the method of acquiring the sample. Future studies should examine this issue using a similar methodology to our study, but including women with benign ndings to al- low calculation of specicity as well as sensitivity. However, it would be reasonable to assume that the improvement in adequate aspirates (achieved with the use of ultrasound-guidance) would not adversely aect specicity. That is, a reduction in insucient aspirates would not be expected to cause false positive cytology. Where a judgement is made to use FNAB in the diagnosis of clinically palpable breast cancer, the use of ultrasound-guidance may improve sensitivity by reduc- ing insucient FNAB aspirates. The evidence from this study should assist in guiding policy and practice in sampling of suspected breast cancer. Acknowledgements Dr Houssami was partly supported by National Health and Medical Research Council grant no. 211205 to the Screening and Test Evaluation Program. We thank all the Florence women who participated in this study. References 1. Ciatto S, Carriagi P, Bulgaresi P, Confortini M, Bonardi R: Fine needle aspiration cytology of the breast: review of 9533 consecutive cases. Breast 2: 8790, 1993 2. Vetrani A, Fulciniti F, Di Benedetto G et al.: Fine-needle aspiration biopsies of breast masses. An additional experience with 1153 cases (19851988) and a meta-analysis. Cancer 69: 736 740, 1992 3. Dixon JM, Anderson TJ, Lamb J, Nixon SJ, Forrest APM: Fine needle aspiration cytology, in relationships to clinical examination and mammography in the diagnosis of a solid breast mass. Br J Surg 71: 593596, 1984 4. Irwig L, Macaskill P, Houssami N: Evidence relevant to the investigation of breast symptoms: the triple test. Breast 11: 215 220, 2002 5. Britton PD: Fine needle aspiration or core biopsy. Breast 8: 14, 1999 N Houssami et al. 58 6. Perry NM, Broeders M, de Wolf C, Thorberg S: European guidelines for quality assurance in mammography screening. Euro- pean Commission Europe Against Cancer, Luxembourg, 2001. 7. NHS Breast Screening Programme Guidelines for Cytology Procedures and Reporting in Breast Cancer Screening. NHSBSP 1992; Publication No. 22 8. Houssami N, Irwig L: Likelihood ratios for clinical examination, mammography, ultrasound and ne needle biopsy in women with breast problems. Breast 7: 8589, 1998 9. Altman DG, Gardner MJ: Statistics with Condence Condence Intervals and Statistical Guidelines, BMJ, London, 1989, pp 3133 10. Zeger SL, Liang K-Y, Albert PS: Models for longitudinal data: a generalized estimating equation approach. Biometrics 44: 1049 1060, 1988 11. Catania S, Ciatto S: Fine-needle aspiration cytology of solid masses. In: Catania S, Ciatto S (eds) Breast Cytology in Clinical Practice. Martin Dunitz, London, 1992, pp 7579 Address for oprints and correspondence: Dr Nehmat Houssami, Screening and Test Evaluation Program, School of Public Health, University of Sydney, NSW 2006, Australia; Tel.: +61-2-9351-5994; Fax: 61-2-9351-5049; E-mail: nehmath@health. usyd.edu.au Ultrasound-guided versus freehand breast needle biopsy 59 Reproducedwith permission of thecopyright owner. Further reproductionprohibited without permission.