Cancer Epidemiology 34 (2010) 457–460

Influence of imaging on touch imprint cytology of breast lesions

Vimal Raj, T. Sivashanmugam, S. Gupta, K. Clarkson, E. Denton, M. Al-Attar *

Department of Radiology, Glenfield Hospital, University Hospitals of Leicester, Groby Road, Leicester LE3 9QP, United Kingdom

A R T I C L E I N F O

Article history:

Accepted 29 April 2010

Available online 27 May 2010

Keywords:

Touch imprint cytology

Breast screening

Breast ultrasound

A B S T R A C T

Purpose: Touch imprint cytology (TIC) facilitates rapid diagnosis of breast diseases in women attending

triple assessment clinics. Some pathologists, in our centre, feel that pathological interpretation of TIC

slides is contentious when the lesions are radiologically indeterminate (R3), as these can lead to

potentially higher false positive or false negative cytology results. We hypothesised that: ‘(R3) lesions

are more likely to have higher false positive or false negative TIC and/or be inadequate for TIC

assessment’. In other words, ‘imaging influences cytological classification especially when indetermi-

nate’. Methods: Review of the data collected in our centre between December 2003 and July 2005. All

patients who attended the one stop symptomatic breast clinic and had a TIC performed following an

ultrasound (US) guided core biopsy (CB) were included. Demographic, radiological, cytological and core

biopsy grading data were collected. Cytology grading was correlated with radiology classification to

assess our hypothesis. Results: A total of 248 patients underwent 254 CB/TIC. The average patient’s age of

the group was 54 years (range of 29–95). On TIC, 186 (73%) were deemed malignant, 23(9%) benign while

33(13%) were inadequate for assessment. There was no false positive or false negative TIC. There was

good correlation between TIC and CB results (p < 0.0001). Thirty-three cases were inadequate (C1) for

cytology assessment, of these16 (48.5%) were indeterminate on imaging. R3 lesions were 6 times more

prone to have C1 cytology (p < 0.0001). Conclusion: Touch imprint cytology is a reliable and efficient

method in running a one stop breast clinic, with the backup of full tissue diagnosis. Careful selection of

cases that would benefit from this technique is highly recommended as a significant number of

radiologically indeterminate lesions are likely to be insufficient for cytological assessment. Further

prospective trials are required to assess this further. Until then the diagnosis in this sub-group should

depend on core biopsy.

� 2010 Elsevier Ltd. All rights reserved.

Contents lists available at ScienceDirect

Cancer EpidemiologyThe International Journal of Cancer Epidemiology, Detection, and Prevention

journal homepage: www.cancerepidemiology.net

1. Introduction

1.1. Background

Breast cancer is the most common malignancy in women,accounting for nearly 30% of all malignancies [1]. Over the lastdecade emphasis has been on providing rapid diagnosis to womenwith suspected breast cancer at a single visit using tripleassessment i.e. clinical examination, imaging and cytology [1–4].

Core biopsy (CB) is used internationally and has bettersensitivity and specificity compared to fine needle aspiration(FNA) [5]. CB, however, does not facilitate same day diagnosis.Touch imprint cytology (TIC) from ultrasound guided CB not onlyallows same day diagnosis but also collects sufficient data requiredfor diagnosis from a single needle test [6]. The technique itself issimple and has been found to rival in accuracy FNA with a highsensitivity (up to 91%) and specificity (up to 89%) [6–10].

* Corresponding author. Tel.: +44 116 2583860; fax: +44 116 2584666.

E-mail address: miaad.al-attar@uhl-tr.nhs.uk (M. Al-Attar).

1877-7821/$ – see front matter � 2010 Elsevier Ltd. All rights reserved.

doi:10.1016/j.canep.2010.04.016

Some pathologists, in our centre, feel that pathologicalinterpretation of TIC slides is contentious when the lesions areradiologically indeterminate (R3). R3 lesions can lead to poten-tially higher false positive or false negative cytology results. Basedon this we hypothesised:

R3 lesions are more likely to:

1. have a higher false positive or false negative TIC and/or;2. be inadequate for TIC assessment.

In other words, ‘imaging influences cytological classificationespecially when indeterminate.’

2. Method

In our practice, a significant number of cases are diagnosedusing FNA cytology with a high sensitivity and specificity. With theincreasing emphasis on the move towards core biopsy, and to keepour practice of complete one stop setup, imprint cytology was theideal technique. We started imprint cytology in December 2003.This technique was performed for ultrasound guided core biopsies

Table 1Comparison of TIC with final CB results.

Core biopsy (CB) total = 254 Touch imprint cytology (TIC)

C5 (n = 165) C4 (n = 21) C3 (n = 12) C2 (n = 23) C1 (n = 33)

B5 (n = 199) 165 17 4 0 13

B4 (n = 0) 0 0 0 0 0

B3 (n = 3) 0 0 1 2 0

B2 (n = 44) 0 4 7 19 14

B1 (n = 8) 0 0 0 2 6

Table 2TIC performance in comparison to the NHSBSP standards.

TIC (%) Preferred (%)

Absolute sensitivity 82.9 >70

Complete sensitivity 93.4 >90

Specificity (full) 43.8 >65

False positive rate 0 <0.5

False negative rate 0 <4

PPV of C5 100 >99

Inadequate rate (overall) 12.9 <15

Suspicious rate (C4) 12.8 <15

V. Raj et al. / Cancer Epidemiology 34 (2010) 457–460458

in clinics supported by a cytology hot reporting system (3 out of 5symptomatic clinics in our centre). We will be referring to theseclinics as the (one stop clinic) in this study. Demand on ourhistopathology service prevented us from providing hot reportingto all our symptomatic clinics. Clinics without this service are theones started later to support service expansion.

All the CB and TIC slides were performed and prepared byconsultant radiologists. Consultant pathologists, using the FNAreporting scheme, assessed the TIC slides in the same clinic [5].Common TIC preparation technique was adopted whereby CBsample was placed on to a dry slide and dragged across it using a21G needle. The slide was then rapidly air-dried and sent to thepathology department via air-tube where it was stained using aRapid Romanowsky Stain and assessed by a consultant pathologistwithin 1 hour. The result was phoned back to the one stop clinic.The final result on the core biopsy tissue was available after twoworking days. All results were discussed at our multidisciplinarymeeting to confirm correlation of clinical, radiological andpathological results. All results were recorded prospectivelycomplying with local audit standards.

2.1. Study design

Data was collected between December 2003 and July 2005. Allpatients who attended the one stop clinic during this period andhad a TIC performed following a US guided CB were included.Demographics, radiological, cytological and core biopsy gradingdata were reviewed. Core biopsy diagnosis was deemed final incases where surgical excision was not performed. TIC results wereassessed against final histology and radiology grading. Studydesign was discussed and accepted by the institution’s ethic’scommittee.

All lesions were radiologically graded on a 1–5 point scale (1:Normal, 2: Benign, 3: Indeterminate, 4: Suspicious, 5: Malignant).The radiological grading was derived from composite of mam-mography and ultrasound appearances. Cytology specimens wereclassified as, C1: Inadequate, C2: Benign, C3: Atypia (not otherwisespecified), C4: Suspicious and C5: Malignant. Cytology grading wascorrelated with radiology classification to assess our hypothesis.TIC results were also compared with final CB and assessed againstthe NHS BSP standards for FNA [5].

2.2. Statistics

TIC classification was compared with imaging and final corebiopsy results to assess correlation and concordance. Specificanalysis was made of C1 group to test our hypothesis. Statisticalanalysis was performed on SPSS for Windows, Rel. 16.0.1.2008.Chicago: SPSS Inc.

Performance indicators for TIC were derived from the NHS BSPguidelines for fine needle aspiration [5]. Using standard defini-tions, positive predictive value, negative predictive value, abso-lute/complete sensitivity and specificity were calculated [11,12].Fisher’s exact test was used to compare categorical data. Spear-man’s rho was used to measure the strength of relationship

between paired ordinal observations. A p value of <0.05 was takenas the threshold for statistical significance.

3. Results

3.1. Patient demographics

In the 20-month period reviewed, 248 (244F:4M) patientsunderwent 254 CB/TIC. Average age of the group was 54 years(range of 29–95). During the same period, 637 ultrasound guidedCB and 2059 FNA procedures were performed in our department.Apart from these, the surgeons performed a few cases of free handcore biopsy and FNA. Imprint cytology was not performed on freehand core biopsy samples.

3.2. TIC performance

Majority of lesions were malignant i.e. 78.3% (199), 17.3% (44)were benign, 3 were indeterminate and 8 were normal breasttissue on final CB diagnoses. On excision biopsy, 2 of the 3indeterminate lesions were borderline phylloides tumour and thethird one was benign sclerosing papillary lesion.

On TIC, 186 (73%) were deemed malignant, 23 (9%) benignwhile 33 (13%) were inadequate for assessment. There was no falsepositive or false negative TIC. There was good correlation betweenTIC and CB results (Grade 2–5) (Spearman’s rho 0.8, p < 0.0001)[Table 1]. Absolute sensitivity of TIC was 83% and the inadequacyrate was 12.5%. Further comparisons with the NHSBSP standardsare shown in Table 2.

3.3. Influence of radiology on TIC

Table 3 compares TIC report against radiological classification.Attention is drawn to the 33 cases that were inadequate (C1) forcytology assessment. Majority of these, 16 (48.5%), were indeter-minate on imaging while 11 (33%) were malignant and 6 (18%)were benign. On the other hand, 95.6% of malignant TIC (C4/5) hadmalignant imaging features (R4/5) and 65.2% of benign TIC hadbenign imaging appearance.

Compared to all other radiological classifications, R3 lesionswere more likely to be inadequate (C1) (Table 4), Fisher’s exact testp < 0.0001, RR = 6.1 (95% CI, 3.4–10.9), odds ratio 10.6 (95% CI, 4.6–24.5).

Table 3TIC compared with radiological classification.

Touch imprint cytology

(TIC) (n = 254)

Radiological findings

R5 R4 R3 R2 R1

C5 (n = 165) 146 13 4 1 1

C4 (n = 21) 16 2 2 1 0

C3 (n = 12) 3 1 5 2 1

C2 (n = 23) 0 1 7 15 0

C1 (n = 33) 10 1 16 6 0

Table 4Distribution of C1 in R3 compared to remainder of radiological classification.

C1 (inadequate) Not C1

R3 16 18

Other than R3 17 203

V. Raj et al. / Cancer Epidemiology 34 (2010) 457–460 459

4. Discussion

To the best of our knowledge this is the only published studyassessing the influence of imaging on cytology results. Our datademonstrate that imaging appearances significantly influenced TICresults. The majority of radiologically indeterminate lesions wereinadequate on TIC.

More than 70% of our patients went home on the same day witha preliminary cytology diagnosis. There were no false positive orfalse negative TIC results in the entire study further strengtheningthe role of TIC in successfully running a one stop breast clinic, withthe backup of full tissue diagnosis.

The designation of an aspirate as ‘inadequate’ is to some extenta subjective matter and may depend on the experience of theinterpreter and/or the aspirator. TIC assessment is reliant on thepresence of sufficient number of epithelial cells for confidentassessment. There are a number of recognised reasons for labellinga smear inadequate. These fall into three main groups: hypocel-lular smear, error in aspiration, spreading or staining and excessiveblood [5]. This study introduces a new cause of C1 result, namely,indeterminate imaging appearances (R3). We could not recogniseany specific cause for this occurrence and believe it might berelated to the inherent nature of these lesions. One cannotoveremphasise the importance of this group of patients, as theywill go home without a provisional diagnosis contrary to theirpreference of receiving the results on the same day [6,13,14].

Thirty-three patients (13%) in our study went home without aprovisional diagnosis, 16 (48%) of these had indeterminate imagingfeatures. From another prospective, 16/34 R3 lesions were C1 (47%)on TIC. Of these 16 cases, 11 were benign, 3 malignant and 2 werenormal breast tissue on core biopsy.

The national policy has been to strive for timely diagnosis forwomen with symptomatic breast diseases and performing TIChelps achieving these goals. All bar one, NHS BSP preferredstandards for FNA were met by TIC in this study. Our full specificitywas 43.8% in comparison to the minimum expected standard of55%. This could be explained by the fact that 32% of the benignlesions were inadequate (C1). If these were taken out of theequation, our full specificity would increase to 63%, well above themedian standards of 58.4 [5].

Only patients attending the one stop symptomatic breast clinicswere included in this study. This is due to the fact that concurrentcytology reporting is available only in these clinics. Although thecytologists reported the TIC slides prospectively, they were notblinded to the imaging classification. We accept criticism for thislack of blinding, however, our C1 rate (13%) is well within thepublished range and the overall TIC results correlate well with the

CB results [5,6,15,16]. A common defined technique was used inpreparing the TIC, reducing the operator bias.

The use of TIC in conjunction with CB also has somedisadvantages. This practice has led to doubling of pathologistworkload as they have to report on two specimens for each caseinstead of the CB alone. To be effective, the TIC reporting isprioritised in this set up thereby impacting on other routinereporting work.

We do not recommend the use of TIC in isolation for makingimportant decisions relating to management. These decisionsshould be made on the histological reports from the CB samplesand following MDT discussion. We believe that TIC acts as anadjuvant and facilitator in running a one stop triple assessmentbreast clinic. The majority of patients could be sent home with aprovisional diagnosis with the full backup of CB. This is the mainreason for the preference of TIC over FNA in our centre. This studyhighlights an important limitation of TIC that every radiologist,pathologist and breast physician must be aware of.

5. Interpretation

Patients with indeterminate imaging appearances are apotential diagnostic challenge and a significant number of thesewill have inadequate TIC assessment. In view of this we proposethat this group of patients should not undergo TIC. This would havea significant financial implications and will save precious time ofthe radiologists and cytopathologists. Patients in this group canalso be sent home without unnecessary delay.

6. Conclusion

Touch imprint cytology allows a reliable and efficient runningof a one stop breast clinic. Careful selection of patients who wouldbenefit from this technique is important as a significant number ofradiologically indeterminate lesions are likely to be inadequate onTIC. Further prospective trials are required to assess thishypothesis further. Until then the diagnosis in this sub-groupshould depend on histology from core biopsy only.

Conflict of interest statement

There are no conflicts of interest, financial disclosures for anyauthor to report.

References

[1] Improving outcomes in breast cancer—manual update. National Institute ofClinical Excellence. Guidance on Cancer Services; 2002.

[2] Labour’s pledge on breast cancer. Department of Health, HMSO; 1997.[3] Breast Surgeons Group of the British Association of Surgical Oncology. Guide-

lines for surgeons in the management of symptomatic breast disease in theUnited Kingdom. Eur J Surg Oncol 1995;21(Suppl. A):1–13.

[4] The Royal College of Radiologists. Cancer care and treatment services; advicefor purchasers and providers. Royal College of Radiologists.

[5] National Health Service Breast Screening Programme. Non-Operative diagnos-tic procedures and reporting. NHSBSP Publication; 2001.

[6] Jones L, Lott MF, Calder CJ, Kutt E. Imprint cytology from ultrasound-guidedcore biopsies: accurate and immediate diagnosis in a one-stop breast clinic.Clin Radiol 2004 Oct;59:903–8.

[7] Carmichael AR, Berresford A, Sami A, Boparai R. Imprint cytology of needlecore-biopsy specimens of breast lesion: is it best of both worlds? Breast(Edinburgh Scotland) 2004;13(June):232–4.

[8] Uematsu T, Kasami M. Core wash cytology of breast lesions by ultrasonogra-phically guided core needle biopsy. Breast Cancer Res Treat 2008;109(May):251–3.

[9] Qureshi NA, Beresford A, Sami S, Boparai R, Gosh S, Carmichael AR. Imprintcytology of needle core-biopsy specimens of breast lesions: is it a usefuladjunct to rapid assessment breast clinics? Breast (Edinburgh Scotland)2007;16(February):81–5.

[10] Hahn PF, Eisenberg PJ, Pitman MB, Gazelle GS, Mueller PR. Cytopathologictouch preparations (imprints) from core needle biopsies: accuracy compared

V. Raj et al. / Cancer Epidemiology 34 (2010) 457–460460

with that of fine-needle aspirates. AJR Am J Roentgenol 1995;165(Novem-ber):1277–9.

[11] Farshid G, Pieterse S. Core imprint cytology of screen-detected breastlesions is predictive of the histologic results. Cancer 2006;108(June):150–6.

[12] Sox Blatt, Higgins Marton. Medical decision-making. Stoneham MA, Butter-worth–Heinmann; 1988.

[13] Harcourt D, Ambler N, Rumsey N, Cawthorn SJ. Evaluation of a one-stop breastlump clinic: a randomized controlled trial. The Breast 1998;7:5.

[14] Harcourt D, Cawthorn SJ, Ambler N, Rumsey N. Psychological impact of sameday diagnosis of symptomatic breast problems: A 2-year follow up. The Breast2000;9(6):351.

[15] Jacobs TW, Silverman JF, Schroeder B, Raza S, Baum JK, Schnitt SJ. Accuracy oftouch imprint cytology of image-directed breast core needle biopsies. ActaCytol 1999;43:169–74.

[16] Khanna AK, Singh MR, Khanna S, Khanna NN. Fine needle aspiration cytology,imprint cytology and tru-cut needle biopsy in breast lumps: a comparativeevaluation. J Indian Med Assoc 1991;89(July):192–5.