Meta-analysis

Meta-analysis of sentinel node imprint cytology in breast cancer

K. Tew1 , L. Irwig2 , A. Matthews1 , P. Crowe3 and P. Macaskill2
1
Breast/Endocrine Surgery and Surgical Oncology Unit, Prince of Wales Hospital, Randwick, 2 Screening and Test Evaluation Program, School of Public
Health, University of Sydney, Sydney and 3 University Department of Surgery, University of New South Wales, Kensington, New South Wales, Australia
Correspondence to: Professor P. Crowe, Breast/Endocrine Surgery and Surgical Oncology Unit, Prince of Wales Hospital, Randwick, New South Wales
2031, Australia (e-mail: kltew@hotmail.com)

Background: Intraoperative diagnosis of breast cancer metastases in axillary sentinel nodes is desirable
to avoid a second operation for lymphadenectomy. Imprint or touch-preparation cytology is a popular
technique that has high specificity and a wide range of sensitivity.
Methods: A systematic search of electronic databases was performed. Included articles were assessed
for methodological and reporting quality. Random-effects model pooled estimates of sensitivity and
specificity were calculated. Single-variable and multivariable meta-regression analyses were performed
for predictors of sensitivity.
Results: Thirty-one studies were included; all were of good methodological quality but reporting
quality varied. Pooled sensitivity of imprint cytology was 63 (95 per cent confidence interval (c.i.) 57
to 69) per cent and specificity was 99 (95 per cent c.i. 98 to 99) per cent. Pooled sensitivity for
macrometastases was 81 per cent and that for micrometastases 22 per cent. Mean or median primary
tumour size (P = 0·004), the prevalence of metastases (P = 0·103) and the proportion of micrometastases
(P = 0·022) were significant risk factors in single-variable meta-regression analysis. Only the proportion
of micrometastases remained significant in multivariable analysis. Frozen sectioning had better sensitivity
than imprint cytology in three of four direct comparisons.
Conclusion: Imprint cytology is simple and rapid, and has good sensitivity for macrometastases. The
significance of poor sensitivity for micrometastases will be determined by trials investigating their natural
history.

Paper accepted 7 June 2005

Published online in Wiley InterScience (www.bjs.co.uk). DOI: 10.1002/bjs.5139

Introduction examination of paraffin-embedded sections, which is the

The practice of sentinel node biopsy in breast cancer is
common among breast surgery units worldwide; when the
sentinel node contains metastases, the current standard
of care is axillary lymphadenectomy. Identification of the
first draining lymph node allows closer examination of
the tissue for the presence of metastatic tumour1 . It also
offers the potential to spare up to 70 per cent of women
with early breast cancer the morbidity of an unnecessary
axillary lymphadenectomy2 . A rapid, reliable intraoperative
method of assessing the sentinel node for the presence of
metastases is desirable so that axillary lymphadenectomy
can be performed at the same operation as the biopsy.
This avoids the morbidity, inconvenience and cost of a
second, separate procedure. In addition, the method of
intraoperative examination should not compromise the
reference standard.
The frequently reported methods of intraoperative
assessment are frozen section histology and imprint, or
touch-preparation, cytology. Reports of frozen section
examination have described a sensitivity of 44–100 per cent
and a specificity close to 100 per cent3 – 9 . However, the
process of freezing, then thawing, the specimen introduces
artefacts and causes the loss of tissue, potentially interfering
with subsequent more detailed pathological examination
of the tissue with paraffin sectioning. Preparation of the
frozen section specimen for examination also takes more
time than imprint cytology.
Imprint cytology is comparable in accuracy to frozen
sectioning10 . It is a simple and rapid method of preparing
a cytology specimen for examination. The excised lymph
node is bisected or sectioned serially. The cut surfaces are

Copyright  2005 British Journal of Surgery Society Ltd British Journal of Surgery 2005; 92: 1068–1080
Published by John Wiley & Sons Ltd
Sentinel node imprint cytology in breast cancer
pressed on to a slide, which is then fixed and stained.
There has been great interest recently in the use of
imprint cytology to examine the sentinel node(s) in women
with breast cancer, but there is variation in the reported
sensitivity of the method. This heterogeneity has been
attributed to several factors: variation in the experience
of pathologists, the use of serial sectioning of the lymph
node, the proportion of lymph node micrometastases in
the series11 and the proportion of metastases that originate
from lobular carcinoma12 .
Sectioning larger sentinel nodes at 2–3-mm intervals,
rather than simply bisecting them, increases the sensitivity
of imprint cytology13 . Micrometastases are tumour
deposits of 2 mm or less in diameter, or detected only
with immunohistochemical stains. They are often not
detected on routine histological examination1 . Lobular
carcinoma metastases resemble benign cells and may be
Explode Breast Neoplasms
11313 articles
AND
Explode
‘Sensitivity and Specificity’
903 articles
AND
Explode Cytodiagnosis
OR Explode Cytological Techniques
OR Explode Cytology
503 articles
Titles and abstracts screened by a single reviewer
39 articles
OR
77 articles
Titles and abstracts screened by a single reviewer, references searched
31 primary studies
Fig. 1 Flow diagram of the literature search
Copyright  2005 British Journal of Surgery Society Ltd
Published by John Wiley & Sons Ltd
1069
scattered throughout a lymph node. These metastases are
also difficult to identify with imprint cytology14 .
The primary objective of this meta-analysis was to review
the published primary studies of imprint cytology and to
calculate a pooled estimate of sensitivity and specificity.
In addition, predictors of sensitivity and specificity were
examined by means of meta-regression. Finally, studies that
compared frozen section histology with imprint cytology
were examined.
Methods
Medline and EMBase databases from 1966 to June 2004
were searched using the strategy shown in Fig. 1. The
‘Explode’ command in Ovid was used to retrieve articles
for that Medical Subject Heading (MeSH) and all of the
more specific subheadings indented beneath it in the MeSH
Explode Lymph Nodes
OR Explode Lymph Node Excision
OR Explode Lymphatic Metastasis
OR Explode Sentinel Lymph Node Biopsy
AND
Touch (textword search)
OR Imprint (textword search)
7803 articles
67 articles
www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
1070
tree structure15 . Textword searching of the same database
range was performed using ‘touch’ or ‘imprint’. A single
reviewer (K.T.) selected the articles and extracted the data
for analysis. References of included articles were searched
for additional studies.
Articles that reported sufficient data for cross-tabulation
of the results of imprint cytology against paraffin sec-
tion examination were included. Trials that assessed
imprint cytology in combination with another intraop-
erative method, such as frozen section examination or
rapid immunohistochemistry, were included. Articles in
languages other than English that were not available in
translated form were not included.
The studies were assessed for the quality of reporting
based on the guidelines set out by the Cochrane Methods
Group and the Standards for Reporting of Diagnostic
Accuracy (STARD) committee16 – 18 .
Data were used on a per patient basis, rather than
per node examined, to calculate sensitivity and specificity,
whenever possible. If there were inconsistencies in
the published data, study authors were contacted for
clarification. Two studies19,20 considered false-positive
results as being true positives on the basis that the cells seen
on cytological examination were a micrometastatic focus
that had been completely removed from the cut surface
by imprinting. These few cases were reclassified as false
positives. Sensitivity and specificity for the subgroups of
sentinel lymph node macrometastases and micrometastases
were calculated, when possible.
Calculation of confidence intervals and random-effects
model (REM) pooled sensitivity and specificity was
performed using the Meta-Test v0·6 program21 . Summary
Table 1 Predictors used in meta-regression
Predictor Outcome type
Year of publication Continuous
Were patients enrolled consecutively? Dichotomous
Was the lymph node sectioned Dichotomous
serially?
Was immunohistochemical staining Dichotomous
used if paraffin sections were
initially negative?
Mean or median primary tumour size Continuous
Prevalence of metastases in each Continuous
study
Proportion of micrometastases in Continuous
each study
Proportion of lobular carcinoma as the Continuous
primary tumour type in each study
Copyright  2005 British Journal of Surgery Society Ltd
Published by John Wiley & Sons Ltd
K. Tew, L. Irwig, A. Matthews, P. Crowe and P. Macaskill
receiver–operator characteristic (SROC) curves were
produced by Meta-Test, using the method described by
Moses et al.22 A statistical test for heterogeneity between
the studies was not performed because the intention was to
examine this with meta-regression.
Meta-regression analysis of sensitivity was performed
using a random-effects logistic regression model fitted
using the SAS statistical analysis package (SAS Insti-
tute, Cary, North Carolina, USA), procedure PROC
NLMIXED. The within-study variability was assumed
to follow a binomial distribution, and the between-study
variability (random effects) was assumed to be normally
distributed23,24 . The predictors were selected before anal-
ysis by agreement (Table 1). If a predictor was not stated
explicitly in a study, that study was excluded from the
analysis. Single-variable analysis was performed for each
predictor, and predictors were considered significant (based
on a t statistic computed as the estimate divided by its
standard error) when the P value did not exceed 0·100.
Multivariable analysis was then performed using the predic-
tors found to be significant in the single-variable analyses.
Results
Literature search
Seventy-seven articles were retrieved from a search of
electronic databases. Thirty-two articles were excluded
because they were not reports of the sensitivity and
specificity of imprint cytology. A further 14 studies were
excluded for the following reasons. Ku48 cross-tabulated
results only for macroscopically negative sentinel nodes;
individual patient data were not available. In the studies of
Expected relationship
Sensitivity increases with publication year, reflecting
increasing experience and/or improving technology
Sensitivity may be overestimated if enrolment was
non-consecutive
Sensitivity increases if the node is serially sectioned
Sensitivity decreases if the study used
immunohistochemical staining because of detection of
micrometastases
Sensitivity increases as tumour size increases, possibly
reflecting the size of metastases
Sensitivity increases as prevalence increases owing to
more experience; alternatively it may decrease
because of more micrometastases
Sensitivity decreases as proportion of micrometastases
increases
Sensitivity decreases as proportion of lobular carcinoma
increases
www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
Sentinel node imprint cytology in breast cancer
Jain et al.49 and Miller et al.50 , patients had locally advanced
tumours and received preoperative chemotherapy. Creager
et al.14 performed a subgroup analysis of lobular carcinoma
alone. Albo et al.51 reported a series of seven men with
breast cancer, which represents about 1 per cent of patients
with breast cancer. Smidt et al.52 examined scrape rather
than touch cytology. Salem et al.53 and Schreiber et al.54
used only immunohistochemical stains on touch imprints.
Pendas et al.55 used ordinary and immunohistochemical
stains but did not provide separate data for each. The
studies of Ahmad et al.56 and Ku et al.57 were available only
in abstract form, from which the sensitivity and specificity
could not be determined. Three studies were not available
in English, so were not consulted. No further studies were
found by searching the references of retrieved articles.
Thirty-one appropriate articles were included in the
meta-analysis. The individual study characteristics are
summarized in Table 2 and their results in Table 3.
Study results and quality appraisal
All of the articles clearly stated the research question, the
population from which patients were drawn and the study
setting. The trials included patients with T stage 0–2 (less
than 5 cm in diameter) primary tumours with clinically
uninvolved axillary lymph nodes. All of the studies were
conducted in hospital-based surgical and pathology units.
Nineteen studies enrolled consecutive patients. Except for
two studies that were non-consecutive, the remainder did
not describe the method of selection. There were 11
prospective studies. Sentinel node status per node examined
was reported by Cohen et al.34 , Yu et al.35 , Lee et al.19 and
Van Diest et al.45 . The rest of the studies reported sentinel
node status per patient. A patient could have more than
one sentinel node, and all identified nodes were removed
for examination.
In eight studies, the sentinel node was bisected
regardless of size. The remaining studies used serial
sectioning. Most of the studies used one of Diff-
Quik (Baxter Diagnostics, McGaw Park, Illinois, USA),
Quickdiff (Baxter, Düdingen, Switzerland), haematoxylin
and eosin or rapid Papanicolaou to stain the touch
preparations. Paraffin sections were universally stained
with haematoxylin and eosin, but ten studies did not
state the number of sections through the node that were
examined. Five studies did not use immunohistochemical
staining on nodes that were negative to haematoxylin and
eosin staining. One study did not describe the protocol for
staining and examination of paraffin sections.
The result of imprint cytology was classified as either
‘positive’ or ‘negative’ in all of the studies. A few used a
Copyright  2005 British Journal of Surgery Society Ltd
Published by John Wiley & Sons Ltd
1071
‘suspicious’ category, which was included in the ‘negative’
category for the purpose of intraoperative diagnosis. The
number of pathologists examining the specimens was stated
in six studies, but the experience of the pathologist(s) was
indicated in only one study.
In all but two of the studies, imprint cytological
examination was intraoperative or blinded to the result of
final histopathological findings. Only three studies stated
that the pathologists who examined the paraffin sections
were blinded to the imprint cytology result. None of the
studies gave details of additional information that was
available to pathologists.
There were nine studies in which the sentinel
lymph node was processed for frozen section exam-
ination after an imprint cytology slide had been
prepared5,7,12,20,29,33,34,44,45 . In five of these stud-
ies7,20,34,44,45 not all of the patients had both techniques
performed; they were excluded from analysis to avoid the
introduction of bias. Independent assessment of the touch
imprint and frozen section methods was performed by
Nagashima et al.5 The same pathologist examined both the
frozen section and the imprint in the study of Sauer et al.29 .
Beach et al.33 and Leidenius et al.12 did not state whether
or not assessment of the frozen section was independent of
the imprint cytology finding.
Confidence intervals were reported in only three
studies31,40,45 . Lee et al.19 used the κ coefficient to evaluate
agreement between imprint cytology and paraffin section
results. No other study used a statistical test to compare
the results of these two techniques.
Patient demographics were reported in 25 studies. The
mean age across all studies was 56 (range 46–60) years.
Mean or median primary tumour size was reported in 18
articles and ranged from 14 to 22 mm. Primary tumour
grade distribution was reported in eight studies.
Noguchi et al.46 confirmed the imprint cytology result
with paraffin section examination in 82 per cent of cases,
but did not specify the reason for its omission in the
remainder. The other studies performed imprint cytology
and paraffin section examination on all patients.
The mean prevalence of sentinel node metastases across
all 31 studies was 34 (range 17–50) per cent. The
distribution of primary tumour characteristics in lymph
node-positive patients was reported in seven studies,
and two studies reported such data for false-negative
procedures. The proportion of invasive lobular carcinoma
was reported in 16 studies, with a mean of 12 (range 4–27)
per cent. The proportion of sentinel node micrometastases
was given in 16 articles, with a mean of 46 (range 26–77)
per cent.
www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
 1072
Table 2 Individual study characteristics

No. of Prospective Consecutive Intraoperative Blinding of

Reference Year patients study cases assessment used Sectioning method Imprint stain Final staining method reader to IC

Ravichandran et al.25 2004 133 NS NS IC Bisected Romanovsky and H&E and IHC NS
May–Grünwald–Giemsa
Zgajnar et al.11 2004 250 NS NS IC 2-mm transverse Hemacolor H&E, then IHC if NS

Published by John Wiley & Sons Ltd

sections negative
Menes et al.26 2003 70 No No IC or FS performed Bisected H&E or Diff-Quik H&E, then IHC if NS
at discretion of negative
pathologist
Aihara et al.27 2003 118 NS Yes IC and IHC 2-mm sections Papanicolaou H&E NS
Nagashima et al.5 2003 124 NS NS IC and FS 2-mm sections Giemsa, Papanicolaou* H&E Yes

Copyright  2005 British Journal of Surgery Society Ltd

Munakata et al.28 2003 69 NS NS IC, IF and IHC 2-mm sections Papanicolaou H&E, then IHC if NS
negative
Sauer et al.29 2003 211 NS NS IC and FS Bisected Diff-Quik NS NS
Leidenius et al.12 2003 375 Yes Yes IC and FS 1–1·5-mm sections Toluidine blue H&E NS
Cserni13 † 2003 55 Yes NS IC 2–3-mm sections if H&E H&E, then IHC if NS
> 5 mm negative
Cserni13 † 2003 55 Yes NS IC Bisected H&E H&E, then IHC if NS
negative
Liang et al.20 2003 20 Yes Yes IC if node < 1 cm 3-mm sections NS H&E, then IHC if NS
in size negative
IC and FS if node
> 1 cm
Mullenix et al.30 2003 71 No Yes IC Bisected if < 5 mm Diff-Quik and H&E H&E, then IHC if NS
2-mm sections if larger negative

www.bjs.co.uk
Karamlou et al.31 2003 142 No NS IC Bisected if node < 8 mm H&E H&E NS
in diameter
Multiple sections if
> 8 mm, thickness NS
Bochner et al.32 2003 53 Yes Yes IC 1–2-mm sections Diff-Quik H&E and IHC NS
Beach et al.33 2003 32 NS NS IC, FS and IHC Bisected or trisected Rapid H&E H&E and IHC NS
Cohen et al.34 2002 84 NS Yes IC, FS and IHC Bisected if node < 1 cm Rapid H&E H&E and IHC No
Trisected or
quadrisected if larger
Yu et al.35 2002 78 NS Yes IC and IHC Bisected or trisected Diff-Quik H&E, then IHC if NS
(size criteria not given) negative
Shiver et al.36 2002 127 No Yes IC Bisected Diff-Quik and H&E H&E, then IHC if NS
negative

(continued overleaf)

British Journal of Surgery 2005; 92: 1068–1080

K. Tew, L. Irwig, A. Matthews, P. Crowe and P. Macaskill
 Table 2 (Continued)

No. of Prospective Consecutive Intraoperative Blinding of

Reference Year patients study cases assessment used Sectioning method Imprint stain Final staining method reader to IC

Published by John Wiley & Sons Ltd

Creager et al.37 ‡ 2002 646 No Yes IC Bisected or 4-mm Diff-Quik , some also had H&E, then IHC if NS
sections H&E negative
Lee et al.19 § 2002 64 NS No IC 1–2-mm sections Diff-Quik (90%) or H&E, then IHC if NS
Papanicolaou negative
Baitchev et al.38 2002 87 NS Yes IC Bisected if node < 1 cm H&E H&E, then IHC if NS
Sentinel node imprint cytology in breast cancer

Larger nodes sectioned negative

Copyright  2005 British Journal of Surgery Society Ltd

at 5-mm intervals
Llatjos et al.39 2002 76 Yes Yes IC 2-mm sections Rapid H&E and IHC Yes
May–Grünwald–Giemsa
Henry-Tillman et al.40 2002 255 No Yes IC, selective FS Bisected if < 2 mm H&E H&E, IHC used NS
2–3-mm sections if selectively
larger
Kane et al.41 2001 150 NS Yes IC Bisected H&E H&E, IHC used NS
selectively
Cserni et al.42 2001 60 NS Yes IC Bisected H&E H&E, then IHC if NS
negative
Teng et al.43 2000 507 Yes Yes IC, selective IHC NS NS H&E and IHC NS
Motomura et al.44 2000 101 Yes Yes IC and FS Bisected if < 4 mm Papanicolaou H&E, then IHC if NS
2-mm sections if larger negative

www.bjs.co.uk
Turner et al.7 1999 278 NS Yes IC and FS Bisected if < 6 mm Diff-Quik H&E, then IHC if NS
Bisected then 2–3-mm negative
sections if larger
Van Diest et al.45 1999 54 Yes Yes IC and FS Bisected if < 10 mm Quickdiff H&E, then IHC if NS
5-mm sections if larger negative
Noguchi et al.46 1999 38 Yes Yes IC, FS and IHC Bisected May–Giemsa H&E, then IHC if NS
negative
Ratanawichitrasin et al.47 1999 55 Yes NS IC Bisected if < 8 mm H&E H&E Yes
3-mm sections if larger

*After a false-positive case during the initial phase, multiple slides and Papanicolaou staining were used. †Two protocols for sectioning sentinel nodes were used in two different groups of patients.
‡May contain some duplicate data from Shiver et al.36 . §Twenty patients in this study had chemotherapy before sentinel node biopsy, but subgroup analysis did not demonstrate any difference in
sensitivity. IC, imprint cytology; NS, not stated; H&E, haematoxylin and eosin; IHC, immunohistochemistry; FS, Frozen section; IF, immunofluorescence.

British Journal of Surgery 2005; 92: 1068–1080

1073
1074 K. Tew, L. Irwig, A. Matthews, P. Crowe and P. Macaskill

Table 3 Individual study results

Mean or Histological Specificity

median age size of Sentinel Sensitivity of of imprint
of patients primary Lobular node Micrometastatic imprint cytology
Reference (years)* tumour (mm)* carcinoma† metastases† disease† cytology (%)‡ (%)‡

Ravichandran et al.25 60 (28–87) Mean 18 NS 53 of 133 (40) NS 86 (71, 94) 97 (90, 99)
Zgajnar et al.11 56 (25–80) Mean 15 (2–67) 17 of 250 (7) 102 of 250 (41) 59 of 102 (58) 34 (25, 44) 99 (95, 100)
Menes et al.26 55 (22–81) Mean 17 (1–40) 11 of 117 (9)§ 20 of 70 (29)§ 11 of 20 (55) 40 (20, 63) 100 (93, 100)
Aihara et al.27 54 (30–81) NS NS 40 of 118 (34) NS 82 (67, 92) 99 (92, 100)
Nagashima et al.5 56 (34–84) Median 16·7 5 of 124 (4) 29 of 124 (23) 8 of 29 (28) 62 (42, 79) 99 (93, 100)
Munakata et al.28 55 (29–86) NS NS 27 of 69 (39)¶ 18 of 37 (49)¶ 74 (54, 88) 100 (92, 100)
Sauer et al.29 NS NS NS 51 of 211 (24) NS 51 (37, 65) 98 (94, 100)
Leidenius et al.12 56 (36–91) NS 102 of 375 (27) 139 of 375 (37) 45 of 139 (32) 68 (59, 75) 99 (97, 100)
Cserni13 NS NS NS 23 of 55 (42) NS 61 (39, 79) 100 (89, 100)
Cserni13 NS NS NS 23 of 55 (42) NS 52 (31, 72) 97 (82, 100)
Liang et al.20 59 (range NS) NS NS 8 of 20 (40) NS 63 (26, 90) 95 (73, 100)
Mullenix et al.30 56 (28–82) Mean 17 (1–56) 7 of 71 (10) 23 of 71 (32) 8 of 23 (35) 48 (28, 69) 100 (93, 100)
Karamlou et al.31 NS NS 36 of 142 (25) 47 of 142 (33) 30 of 47 (64) 70 (55, 82) 100 (96, 100)
Bochner et al.32 58 (26–86) NS NS 17 of 53 (32) 7 of 17 (41) 53 (29, 76) 100 (90, 100)
Beach et al.33 56 (32–86) Mean 19 (5–65) 3 of 32 (9) 13 of 32 (41)¶ 6 of 14 (43)¶ 69 (39, 90) 100 (82, 100)
Cohen et al.34 55 (36–76) Mean 17 (8–80) 9 of 91 (10)# 35 of 91 (38)# 27 of 35 (77) 37 (22, 55) 100, (85, 100)
Yu et al.35 46 (26–82) NS NS 50 of 189 (26)# NS 95 (75, 100)** 98 (89, 100)**
Shiver et al.36 NS NS 15 of 127 (12) 36 of 127 (28)¶ 18 of 41 (44)¶ 53 (36, 69) 100 (96, 100)
Creager et al.37 58 (21–88) Mean 17 (1–88) 61 of 646 (9) 204 of 646 (32) 97 of 204 (48) 53 (46, 60) 98 (97, 99)
Lee et al.19 56 (30–85) Mean 18 (2–56) NS 17 of 64 (27)¶ 12 of 32 (38)¶ 71 (44, 89) 94 (71, 100)
Baitchev et al.38 54 (28–73) Median 22 (8–37) NS 25 of 87 (29) NS 83 (68, 92) 97 (88, 99)
Llatjos et al.39 57 (32–85) Mean 18 (4–43) 3 of 76 (4) 31 of 76 (41) 8 of 31 (26) 68 (49, 83) 100 (92, 100)
Henry-Tillman et al.40 58 (32–84) Mean 19 25 of 255 (10) 43 of 255 (17) NS 94 (85, 98) 99 (98, 100)
Kane et al.41 57 (34–75) Median 14 (2–47) 27 of 152 (18)†† 37 of 150 (25)†† NS 41 (24, 61) 100 (97, 100)
Cserni42 59 (range NS) Mean 22 (1–60) 4 of 60 (7) 30 of 60 (50) 12 of 30 (40) 63 (44, 79) 100 (88, 100)
Teng et al.43 NS NS 65 of 507 (13)‡‡ 182 of 519 (35)‡‡ NS 62 (54, 69) 100 (99, 100)
Motomura et al.44 51 (28–75) Median 21 (7–50) NS 33 of 101 (33) NS 91 (75, 98) 99 (91, 100)
Turner et al.7 57 (33–89) Median 14 (1–62) 27 of 278 (10) 111 of 278 (40) 64 of 111 (58) 48 (38, 57)§§ 100 (98, 100)
Van Diest et al.45 54 (30–72) Mean 19 NS 23 of 54 (43) NS 64 (41, 82) 100 (85, 100)
Noguchi et al.46 54 (28–82) NS NS 13 of 38 (34) NS 46 (21, 74) 100 (86, 100)
Ratanawichitrasin et al.47 56 (range NS) Mean 17 NS 15 of 55 (27) NS 93 (66, 100) 100 (91, 100)

Values in parentheses are *ranges, †percentages and ‡95 per cent confidence intervals. §A series of 117 patients, 11 of whom had lobular carcinoma; 70
patients had imprint cytology but the number with lobular carcinoma in this group was not given, hence the difference in denominators. ¶Sentinel node
status per patient and micrometastases per node. #Data per node reported. **Values incorporate the results of imprint immunohistochemistry for six cases
that were negative on cytology; sensitivity in the worst-case scenario would be 68 per cent. ††Excluded two patients from analysis because of a break in
study protocol, but did not state whether or not they had lobular carcinoma. ‡‡Twelve patients had bilateral tumours and sentinel node biopsy.
§§Combined value for imprint cytology and frozen section; separate data for imprint cytology not provided. NS, not stated.

In all of the studies, cross-tabulation of imprint cytology
against the paraffin section result was either presented or
could be determined from the data given. There were no
missing results.
was evident for sensitivity, but not for specificity. The
REM pooled estimate of sensitivity was 63 (95 per cent
confidence interval (c.i.) 57 to 69) per cent, and that
for specificity was 99 (95 per cent c.i. 98 to 99) per cent
(Fig. 3).

Sensitivity and specificity

The total number of patients with positive sentinel nodes
on final histopathological examination was 1343. The
total number with negative sentinel nodes was 2862.
Sensitivity ranged from 34 to 95 per cent, with most
studies in the range of 50–70 per cent (Fig. 2). Specificity
ranged from 94 to 100 per cent. The area under the
weighted SROC curve was 0·98 (Fig. 2). Heterogeneity
Results for macrometastases and micrometastases
Data were available for 336 patients with macrometastases
and 336 with micrometastases (Table 4). There were
1164 patients with no metastases. The reported values
of sensitivity for macrometastases ranged from 70 to
98 per cent. The pooled sensitivity was 81 (95 per cent
c.i. 74 to 86) per cent. For micrometastases, sensitivity

Copyright  2005 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
Published by John Wiley & Sons Ltd
Sentinel node imprint cytology in breast cancer 1075

100 100
90
90
80

80 70
60
70

%
50
40
60
30
Sensitivity

50 20
10
40 0
Sensitivity Specificity
30
Fig. 3Plot of sensitivity and specificity of imprint cytology.
20 Values are random-effects model pooled estimates with
95 per cent confidence intervals and ranges
10

calculated according to the method of Diamond58 . The

0 10 20 30 40 50 60 70 80 90 100
1 −Specificity
REM pooled estimate of false negatives resulting from
failure to detect micrometastases was 82 (95 per cent c.i.
Fig. 2Summary receiver–operator characteristic curve for 75 to 88) per cent (Table 4). False positives occurred when
imprint cytology. The sizes of the ellipses are inversely epithelial histiocytes, germinal centre lymphocytes and
proportional to study variance. The cross (×) represents the activated endothelial cells were mistaken for tumour cells7 .
random-effects model pooled estimate of sensitivity and A further possibility was contamination of the specimen
specificity from another sample when it was immersed in the fixing
solution11 .

ranged from 5 to 57 per cent, with a pooled estimate of 22

(95 per cent c.i. 14 to 33) per cent. These results are shown
in Fig. 4, which illustrates the large difference between
these estimates of sensitivity. The specificity ranged from
94 to 100 per cent; pooled specificity was 99 (95 per cent
c.i. 98 to 99) per cent.
The proportion of false negatives due to micrometastases
in each study is presented in Table 4, with 95 per cent c.i.
Comparison of imprint cytology and frozen section
examination
In three of the studies5,12,29 , frozen sectioning performed
better than imprint cytology (Table 5 and Fig. 5). In the
report by Leidenius et al.12 , the confidence intervals did
not overlap. The pooled sensitivity and specificity for
imprint cytology were 62 (95 per cent c.i. 53 to 70) and

Table 4 Subset of 11 studies from which sensitivity for macrometastases and micrometastases could be calculated

False
Metastases Sensitivity for Sensitivity for negatives due to
Reference (%) Micrometastases* macrometastases† micrometastases† micrometastases†

Zgajnar et al.11 41 59 of 102 (58) 74 (59, 86) 5 (1, 15) 84 (72, 91)
Menes et al.26 29 11 of 20 (55) 78 (40, 96) 8 (0, 40) NS
Munakata et al.28 39 18 of 37 (49) 70 (53, 84) 39 (18, 64) NS
Karamlou et al.31 33 30 of 47 (64) 94 (69, 100) 57 (38, 74) 93 (64, 100)
Bochner et al.32 32 7 of 17 (41) 70 (36, 92) 29 (5, 69) 88 (47, 99)
Shiver et al.36 28 18 of 41 (44) 87 (65, 97) 22 (7, 48) 82 (56, 95)
Creager et al.37 32 97 of 204 (48) 82 (73, 89) 22 (14, 31) 80 (70, 87)
Lee et al.19 26 12 of 32 (38) 75 (51, 90) 25 (7, 57) 64 (36, 86)
Llatjos et al.39 41 8 of 31 (26) 91 (70, 98) 11 (1, 49) 80 (44, 96)
Cserni42 50 12 of 30 (40) 85 (61, 96) 20 (4, 56) 77 (46, 94)
Turner et al.7 40 64 of 111 (58) 98 (87, 100) 11 (5, 22) 98 (90, 100)
Pooled (REM) estimate 81 (74, 86) 22 (14, 33) 82 (75, 88)

Values in parentheses are *percentages and †95 per cent confidence intervals. NS, not stated; REM, random-effects model.

Copyright  2005 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
Published by John Wiley & Sons Ltd
1076 K. Tew, L. Irwig, A. Matthews, P. Crowe and P. Macaskill

100 However, it was decided to model only the variation in

90 sensitivity with random-effects logistic regression. This
80 was an approximation that should not be misleading, given
70 the data.
Sensitivity (%)

60
50
40 Single-variable analysis
30
The statistically significant predictors of axillary node
20
involvement were the mean or median primary tumour size
10
(t = 3·28, 17 d.f., P = 0·004), the prevalence of metastasis
0
Macrometastases Micrometastases (t = −1·68, 30 d.f., P = 0·103) and the proportion of
micrometastases (t = −2·54, 15 d.f., P = 0·022). According
Comparison of the sensitivity of imprint cytology for
Fig. 4
to guidelines of Sterne and Smith66 , we felt that the P values
macrometastases and micrometastases. Values are
random-effects model pooled estimates with 95 per cent for prevalence of metastasis could be considered significant.
confidence intervals and ranges Sensitivity increased as mean tumour size increased.
Sensitivity decreased as both the prevalence of metastases
and the proportion of micrometastases increased. Values
Nagashima et al.5 (n = 124) of sensitivity over the interquartile range of the predictors
Sauer et al.29 (n = 214) were calculated from each single-variable model (Table 6).
Leidenius et al.12 (n = 375) The estimated variance of the random effects differed
Beach et al.33 (n = 32) significantly from zero in models for tumour size (t = 2·04,
0 20 40 60 80 100
17 d.f., P = 0·056) and prevalence of metastases (t = 2·86,
Imprint cytology Sensitivity (%) 30 d.f., P = 0·008), providing evidence of unexplained
Frozen section heterogeneity among the studies. This was not the case
with the model for micrometastases (t = 1·35, 15 d.f.,
Fig. 5Comparison of the sensitivity of imprint cytology and P = 0·199).
frozen section examination. The direction of the arrow shows the
change when going from imprint cytology to frozen sectioning
Multivariable analysis
99 (95 per cent c.i. 97 to 99) per cent respectively. The It was proposed that the effect of both prevalence of
respective values for frozen sectioning were 76 (95 per cent metastases and tumour size could be confounded by
c.i. 65 to 84) and 99 (95 per cent c.i. 98 to 100) per cent. the proportion of micrometastases. For example, a study
with a relatively high prevalence of metastases may
have a high proportion of micrometastases, leading to
Meta-regression
poor sensitivity. A study with larger mean tumour size
The SROC curve (Fig. 2) showed that there was little may have a smaller proportion of micrometastases and
variation in specificity across the studies. The authors better sensitivity. There was no a priori reason to expect
expected a complex, curvilinear relationship between confounding between prevalence of metastases and mean
sensitivity and specificity, as described by Moses et al.22 . tumour size. Multivariable meta-regression was undertaken

Table 5 Comparison of imprint cytology and frozen sectioning

Imprint cytology Frozen section

Reference n Sensitivity (%) Specificity (%) Sensitivity (%) Specificity (%)

5
Nagashima et al. 124 62 (42, 79) 99 (93, 100) 79 (60, 91) 100 (96, 100)
Sauer et al.29 214 51 (37, 65) 98 (94, 100) 75 (60, 85) 99 (96, 100)
Leidenius et al.12 375 68 (59, 75) 99 (97, 100) 83 (76, 89) 99 (97, 100)
Beach et al.33 32 69 (39, 90) 100 (82, 100) 54 (26, 79) 100 (82, 100)
Pooled REM estimate 62 (53, 70) 99 (97, 99) 76 (65, 84) 99 (98, 100)

Values in parentheses are 95 per cent confidence intervals. REM, random-effects model.

Copyright  2005 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
Published by John Wiley & Sons Ltd
Sentinel node imprint cytology in breast cancer 1077

Table 6 Single-variable model of imprint cytology sensitivity by axillary lymphadenectomy would be uncommon with
predictor a pooled specificity of 99 per cent. Therefore, about
21 per cent of all patients having a sentinel node biopsy
Sensitivity Absolute change
Predictor (%) (%) would be correctly identified as having metastases and
proceed to axillary lymphadenectomy within the same
Mean or median tumour size (mm)*
operation.
14 42·1 —
16 55·1 + 13·0 Menes et al.26 stated that studies of imprint cytology
18 67·5 + 12·4 could not be compared because of differences between
20 77·8 + 10·3 patients, imprint sectioning and staining, protocol of
22 85·6 + 7·8
Prevalence of metastases (%)*
haematoxylin and eosin-stained paraffin section examina-
17.0 78·0 — tion and pathologist experience. The results of the present
25.3 72·4 − 5·6 review were constrained by the quality of the reporting in
33.5 66·1 − 6·3
the articles. Descriptions of the imprint cytology proce-
41.8 59·1 − 7·0
50.0 51·8 − 7·3 dure were very good. The reference standard could have
Proportion of micrometastases (%)* been described better in many cases; sensitivity may be
26.0 66·7 — overestimated if rigorous examination for micrometastases
38.8 60·4 − 6·3
51.5 53·7 − 6·7
was not performed. Blinding to the reference standard
64.3 46·9 − 6·8 result would have occurred in most studies. Blinding to
77.0 40·2 − 6·7 the imprint cytology result and other relevant information
such as tumour size, grade and type was poorly described.
*Values represent the minimum, 25 per cent, 50 per cent, 75 per cent and
Lack of blinding has been associated with overestimation
maximum respectively over the interquartile range of each predictor.
of test performance59 . The experience of the pathologists
reviewing the imprints was often not described. Selection
to determine whether there was evidence for such an of patients was not well described. Bias may have occurred
interaction. if sentinel node biopsy was performed in patients with a
When the prevalence of metastases and the proportion low suspicion of having axillary disease. Verification bias
of micrometastases were included in a model (data available was unlikely as almost all patients had both tests. There
for 16 studies), the coefficient for prevalence of metastases were 15 studies that did not present results in a 2 × 2 table,
became non-significant (t = 0·22, 15 d.f., P = 0·829), a number that was surprising.
whereas that for the proportion of micrometastases Sensitivity increased as mean or median tumour size
remained significant (t = −2·54, 15 d.f., P = 0·023). The in a study increased. This could be explained by larger
coefficient for micrometastases changed from −2·142 to primary tumours being associated with larger metastases.
−2·174, a difference of 1·5 per cent. Hence, the addition of Zgajnar et al.11 found that the rate of immediate axillary
prevalence of metastases to the model did not greatly alter lymphadenectomy for tumours smaller than 10 mm was
the effect of the proportion of micrometastases. There was 5 per cent, compared with 18 per cent for tumours larger
evidence for confounding between these two predictors. than 10 mm, a difference that was statistically significant.
When both size and micrometastases were included in Bochner et al.60 reported similar findings. Cohen et al.34
the model, the proportion of micrometastases remained reviewed this in their discussion and found five reports
significant (t = −2·95, 10 d.f., P = 0·015), but tumour size of a positive correlation between the size of the tumour
did not (t = 1·23, 10 d.f., P = 0·247). Again, there was and that of metastases. It was expected that there would
evidence for confounding. be a smaller proportion of micrometastases in studies
with larger mean tumour size, but multivariable analysis
showed that the effect of the proportion of micrometastases
Discussion
was predominant. Further analysis using individual patient
The initial reports of imprint cytology showed a high data would be more meaningful, rather than inferring the
degree of accuracy, especially the series of Motomura relationship by using the mean from each study.
et al.44 and Ratanawichitrasin et al.47 . Subsequent reports Sensitivity decreased as the prevalence of metastases and
have not been as encouraging. In the present meta-analysis the proportion of micrometastases increased. However,
a 63 per cent pooled sensitivity for imprint cytology and a the coefficient of prevalence became non-significant in
34 per cent mean prevalence of metastases were calculated multivariable analysis. There was confounding between
across 31 studies. False positives resulting in an unnecessary these two factors, with the proportion of micrometastases

Copyright  2005 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
Published by John Wiley & Sons Ltd
1078
being more influential. Despite low overall sensitivity,
imprint cytology is very accurate when macrometastases
are considered.
The consequences of not detecting micrometastatic
disease in early breast cancer remain unknown. There
are grounds for not performing axillary lymphadenectomy
in these patients. Chu et al.61 showed that the size of the
primary tumour and the size of sentinel node metastases are
predictors of non-sentinel node metastasis. Furthermore,
the rate of non-sentinel node involvement was greater in
patients who had macrometastases. Turner et al.7 found
that patients with a positive finding on imprint cytology
were more likely to have positive non-sentinel lymph nodes
elsewhere in the axilla. Grube and Giuliano62 described
five reasons for not performing axillary lymphadenectomy
when micrometastatic disease is found: the sentinel lymph
node is often the only node with metastatic disease;
the decision to give adjuvant chemotherapy is often
based on primary tumour- and patient-specific factors;
radiation given with breast-conserving surgery extends to
the axilla; axillary node metastases are predictors, but not
determinants of survival; and few studies have included a
‘watch-and-wait’ approach to the axilla.
Contrary to the above arguments would be an obser-
vation of survival advantage from axillary lymphadenec-
tomy even when the nodes were negative on pathological
examination63 . Survival is correlated with the number of
lymph nodes removed, but is this the effect of micrometas-
tases? Both the Ludwig Breast Cancer Study Group64
and Clare et al.65 found retrospectively that progno-
sis was worse when micrometastases were present. The
National Surgical Breast and Bowel Project NSABP-32
trial in progress addresses the issue of whether adding
lymphadenectomy to sentinel node biopsy, regardless of
sentinel node status, is associated with better survival
outcome63 .
This review found that imprint cytology did not compare
favourably with frozen section examination; however, the
impact of tissue loss when processing the frozen section
specimen could not be assessed. This might account
for the seemingly better sensitivity of imprint cytology,
because micrometastases are lost to examination. The
importance of detecting micrometastases in determining
prognosis makes frozen section examination unattractive
for intraoperative diagnosis.
Future reports of imprint cytology or other intraopera-
tive assessment of sentinel nodes should aim for better study
design and quality of reporting. Several reports involved
the use of rapid immunohistochemical19,20,27,33,34,53 or
immunofluorescence28 stains, with results superior to those
Copyright  2005 British Journal of Surgery Society Ltd
Published by John Wiley & Sons Ltd
K. Tew, L. Irwig, A. Matthews, P. Crowe and P. Macaskill
of ordinary stains. However, use of these stains is con-
strained by cost, availability and the time taken to prepare
the specimen. False positives occur with extrafollicular
dendritic cells that also stain for cytokeratin.
Acknowledgements
The authors thank Dr J. Lau (New England Medical
Centre, Boston, Massachusetts, USA) for use of Meta-Test
v0·6 software, and Ms Siew F. Chan (Screening and Test
Evaluation Program, School of Public Health, University
of Sydney) for preparing Fig. 5. This work was supported
in part by National Health and Medical Research Council
Program Grant No. 211205 to the Screening and Test
Evaluation Program.
References
1 Giuliano AE, Dale PS, Turner RR, Morton DL, Evans SW,
Krasne DL. Improved axillary staging of breast cancer with
sentinel lymphadenectomy. Ann Surg 1995; 222: 394–399.
2 Miltenburg DM, Miller C, Karamlou TB, Brunicardi FC.
Meta-analysis of sentinel lymph node biopsy in breast cancer.
J Surg Res 1999; 84: 138–142.
3 Flett MM, Going JJ, Stanton PD, Cooke TG. Sentinel node
localization in patients with breast cancer. Br J Surg 1998;
85: 991–993.
4 Gulec SA, Su J, O’Leary JP, Stolier A. Clinical utility of
frozen section in sentinel node biopsy in breast cancer. Am
Surg 2001; 67: 529–532.
5 Nagashima T, Suzuki M, Yagata H, Nikaido T, Horiuchi F,
Koda K et al. Intraoperative cytologic diagnosis of sentinel
node metastases in breast cancer. Acta Cytol 2003; 47:
1028–1032.
6 Tanis PJ, Boom RP, Koops HS, Faneyte IF, Peterse JL,
Nieweg OE et al. Frozen section investigation of the sentinel
node in malignant melanoma and breast cancer. Ann Surg
Oncol 2001; 8: 222–226.
7 Turner RR, Hansen NM, Stern SL, Giuliano AE.
Intraoperative examination of the sentinel lymph node for
breast carcinoma staging. Am J Clin Pathol 1999; 112:
627–634.
8 Veronesi U, Paganelli G, Galimberti V, Viale G, Zurrida S,
Bedoni M et al. Sentinel-node biopsy to avoid axillary
dissection in breast cancer with clinically negative
lymph-nodes. Lancet 1997; 349: 1864–1867.
9 Weiser MR, Montgomery LL, Susnik B, Tan LK,
Borgen PI, Cody HS. Is routine intraoperative
frozen-section examination of sentinel lymph nodes in breast
cancer worthwhile? Ann Surg Oncol 2000; 7: 651–655.
10 Scucchi LF, Di Stefano D, Cosentino L, Vecchione A. Value
of cytology as an adjunctive intraoperative diagnostic
method. An audit of 2250 consecutive cases. Acta Cytol 1997;
41: 1489–1496.
www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
Sentinel node imprint cytology in breast cancer
11 Zgajnar J, Frkovic-Grazio S, Besic N, Hocevar M,
Vidergar-Kralj B, Gerljevic A et al. Low sensitivity of the
touch imprint cytology of the sentinel lymph node in breast
cancer patients – results of a large series. J Surg Oncol 2004;
85: 82–86.
12 Leidenius MH, Krogerus LA, Toivonen TS, Von
Smitten KJ. The feasibility of intraoperative diagnosis of
sentinel lymph node metastases in breast cancer. J Surg Oncol
2003; 84: 68–73.
13 Cserni G. Effect of increasing the surface sampled by imprint
cytology on the intraoperative assessment of axillary sentinel
lymph nodes in breast cancer patients. Am Surg 2003; 69:
419–423.
14 Creager AJ, Geisinger KR, Perrier ND, Shen P, Shaw JA,
Young PR et al. Intraoperative imprint cytologic evaluation
of sentinel lymph nodes for lobular carcinoma of the breast.
Ann Surg 2004; 239: 61–66.
15 Greenhalgh T. How to read a paper: the Medline database.
BMJ 1997; 315: 180–183.
16 Cochrane methods groups on systematic review of screening
and diagnostic tests: recommended methods, updated 6 June
1996. http://www.cochrane.org/cochrane/sadtdoc1.htm 1996
[accessed 15 March 2004]
17 Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA,
Glasziou PP, Irwig LM et al. The STARD statement for
reporting studies of diagnostic accuracy: explanation and
elaboration. Ann Intern Med 2003; 138: W1–W12.
18 Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA,
Glasziou PP, Irwig LM et al. Towards complete and accurate
reporting of studies of diagnostic accuracy: the STARD
Initiative. Ann Intern Med 2003; 138: 40–44.
19 Lee A, Krishnamurthy S, Sahin A, Symmans WF, Hunt K,
Sneige N. Intraoperative touch imprint of sentinel lymph
nodes in breast carcinoma patients. Cancer 2002; 96:
225–231.
20 Liang R, Craik J, Juhasz ES, Harman CR. Imprint cytology
versus frozen section: intraoperative analysis of sentinel lymph
nodes in breast cancer. ANZ J Surg 2003; 73: 597–599.
21 Lau J. Meta-Test version 0.6. New England Medical Centre:
Boston, 1997.
22 Moses LE, Shapiro D, Littenberg B. Combining
independent studies of a diagnostic test into a summary ROC
curve: data-analytic approaches and some additional
considerations. Stat Med 1993; 12: 1293–1316.
23 Macaskill P. Empirical Bayes estimates generated in a
hierarchical summary ROC analysis agreed closely with those
of a full Bayesian analysis. J Clin Epidemiol 2004; 57:
925–932.
24 van Houwelingen HC, Arends LR, Stijnen T. Advanced
methods in meta-analysis: multivariate approach and
meta-regression. Stat Med 2002; 21: 589–624.
25 Ravichandran D, Kocjan G, Falzon M, Ball RY, Ralphs DN.
Imprint cytology of the sentinel lymph node in the
assessment of axillary node status in breast carcinoma. Eur J
Surg Oncol 2004; 30: 238–242.
Copyright  2005 British Journal of Surgery Society Ltd
Published by John Wiley & Sons Ltd
1079
26 Menes TS, Tartter PI, Mizrachi H, Smith SR, Estabrook A.
Touch preparation or frozen section for intraoperative
detection of sentinel lymph node metastases from breast
cancer. Ann Surg Oncol 2003; 10: 1166–1170.
27 Aihara T, Munakata S, Morino H, Takatsuka Y. Touch
imprint cytology and immunohistochemistry for the
assessment of sentinel lymph nodes in patients with breast
cancer. Eur J Surg Oncol 2003; 29: 845–848.
28 Munakata S, Aihara T, Morino H, Takatsuka Y. Application
of immunofluorescence for intraoperative evaluation of
sentinel lymph nodes in patients with breast carcinoma.
Cancer 2003; 98: 1562–1568.
29 Sauer T, Engh V, Holck AM, Sorpebol G, Heim M, Furu I
et al. Imprint cytology of sentinel lymph nodes in breast
cancer. Experience with rapid, intraoperative diagnosis and
primary screening by cytotechnologists. Acta Cytol 2003; 47:
768–773.
30 Mullenix PS, Carter PL, Martin MJ, Steele SR, Scott CL,
Walts MJ et al. Predictive value of intraoperative touch
preparation analysis of sentinel lymph nodes for axillary
metastasis in breast cancer. Am J Surg 2003; 185: 420–424.
31 Karamlou T, Johnson NM, Chan B, Franzini D, Mahin D.
Accuracy of intraoperative touch imprint cytologic analysis of
sentinel lymph nodes in breast cancer. Am J Surg 2003; 185:
425–428.
32 Bochner MA, Farshid G, Dodd TJ, Kollias J, Gill PG.
Intraoperative imprint cytologic assessment of the sentinel
node for early breast cancer. World J Surg 2003; 27: 430–432.
33 Beach RA, Lawson D, Waldrop SM, Cohen C. Rapid
immunohistochemistry for cytokeratin in the intraoperative
evaluation of sentinel lymph nodes for metastatic breast
carcinoma. Appl Immunohistochem Mol Morphol 2003; 11:
45–50.
34 Cohen C, Alazraki N, Styblo T, Waldrop SM, Grant SF,
Larsen T. Immunohistochemical evaluation of sentinel
lymph nodes in breast carcinoma patients. Appl
Immunohistochem Mol Morphol 2002; 10: 296–303.
35 Yu JC, Hsu GC, Liu YC, Sheu LF, Li SH, Chao TY.
Sentinel node biopsy in early breast cancer in Taiwan. World
J Surg 2002; 26: 1365–1369.
36 Shiver SA, Creager AJ, Geisinger K, Perrier ND, Shen P,
Levine EA. Intraoperative analysis of sentinel lymph nodes
by imprint cytology for cancer of the breast. Am J Surg 2002;
184: 424–427.
37 Creager AJ, Geisinger KR, Shiver SA, Perrier ND, Shen P,
Shaw JA et al. Intraoperative evaluation of sentinel lymph
nodes for metastatic breast carcinoma by imprint cytology.
Mod Pathol 2002; 15: 1140–1147.
38 Baitchev G, Gortchev G, Todorova A. Intraoperative
sentinel lymph node examination by imprint cytology during
breast surgery. Curr Med Res Opin 2002; 18: 185–187.
39 Llatjos M, Castella E, Fraile M, Rull M, Julian FJ, Fuste F
et al. Intraoperative assessment of sentinel lymph nodes in
patients with breast carcinoma: accuracy of rapid imprint
cytology compared with definitive histologic workup. Cancer
2002; 96: 150–156.
www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080
1080
40 Henry-Tillman RS, Korourian S, Rubio IT, Johnson AT,
Mancino AT, Massol N et al. Intraoperative touch
preparation for sentinel lymph node biopsy: a 4-year
experience. Ann Surg Oncol 2002; 9: 333–339.
41 Kane JM III, Edge SB, Winston JS, Watroba N, Hurd TC.
Intraoperative pathologic evaluation of a breast cancer
sentinel lymph node biopsy as a determinant for synchronous
axillary lymph node dissection. Ann Surg Oncol 2001; 8:
361–367.
42 Cserni G. The potential value of intraoperative imprint
cytology of axillary sentinel lymph nodes in breast cancer
patients. Am Surg 2001; 67: 86–91.
43 Teng S, Dupont E, McCann C, Wang J, Bolano M,
Durand K et al. Do cytokeratin-positive-only sentinel lymph
nodes warrant complete axillary lymph node dissection in
patients with invasive breast cancer? Am Surg 2000; 66:
574–578.
44 Motomura K, Inaji H, Komoike Y, Kasugai T, Nagumo S,
Noguchi S et al. Intraoperative sentinel lymph node
examination by imprint cytology and frozen sectioning
during breast surgery. Br J Surg 2000; 87: 597–601.
45 Van Diest PJ, Torrenga H, Borgstein PJ, Pijpers R,
Bleichrodt RP, Rahusen FD et al. Reliability of intraoperative
frozen section and imprint cytological investigation of
sentinel lymph nodes in breast cancer. Histopathology 1999;
35: 14–18.
46 Noguchi M, Bando E, Tsugawa K, Miwa K, Yokoyama K,
Nakajima K et al. Staging efficacy of breast cancer with
sentinel lymphadenectomy. Breast Cancer Res Treat 1999; 57:
221–229.
47 Ratanawichitrasin A, Biscotti CV, Levy L, Crowe JP. Touch
imprint cytological analysis of sentinel lymph nodes for
detecting axillary metastases in patients with breast cancer. Br
J Surg 1999; 86: 1346–1348.
48 Ku NN. Pathologic examination of sentinel lymph nodes in
breast cancer. Surg Oncol Clin N Am 1999; 8: 469–479.
49 Jain P, Kumar R, Anand M, Asthana S, Deo SV, Gupta R
et al. Touch imprint cytology of axillary lymph nodes after
neoadjuvant chemotherapy in patients with breast carcinoma.
Cancer 2003; 99: 346–351.
50 Miller AR, Thomason VE, Yeh IT, Alrahwan A, Sharkey FE,
Stauffer J et al. Analysis of sentinel lymph node mapping with
immediate pathologic review in patients receiving
preoperative chemotherapy for breast carcinoma. Ann Surg
Oncol 2002; 9: 243–247.
51 Albo D, Ames FC, Hunt KK, Ross MI, Singletary SE,
Kuerer HM. Evaluation of lymph node status in male breast
cancer patients: a role for sentinel lymph node biopsy. Breast
Cancer Res Treat 2003; 77: 9–14.
52 Smidt ML, Besseling R, Wauters CA, Strobbe LJ.
Intraoperative scrape cytology of the sentinel lymph node in
patients with breast cancer. Br J Surg 2002; 89: 1290–1293.
Copyright  2005 British Journal of Surgery Society Ltd
Published by John Wiley & Sons Ltd
K. Tew, L. Irwig, A. Matthews, P. Crowe and P. Macaskill
53 Salem AA, Douglas-Jones AG, Sweetland HM,
Newcombe RG, Mansel RE. Evaluation of axillary lymph
nodes using touch imprint cytology and
immunohistochemistry. Br J Surg 2002; 89: 1386–1389.
54 Schreiber RH, Pendas S, Ku NN, Reintgen DS, Shons AR,
Berman C et al. Microstaging of breast cancer patients using
cytokeratin staining of the sentinel lymph node. Ann Surg
Oncol 1999; 6: 95–101.
55 Pendas S, Dauway E, Cox CE, Giuliano R, Ku NN,
Schreiber RH et al. Sentinel node biopsy and cytokeratin
staining for the accurate staging of 478 breast cancer patients.
Am Surg 1999; 65: 500–505.
56 Ahmad N, Ku NN, Nicosia SV, Smith PV,
Muro-Cacho CA, Livingston S et al. Evaluation of sentinel
lymph node imprints in breast cancer: role of intraoperative
cytokeratins immunostaining in breast cancer staging. Acta
Cytol 1998; 42: 1218 (Abstract).
57 Ku NN, Ahmad N, Smith PV, Cox CE, Shons A,
Reintgen DS et al. Intraoperative imprint cytology of sentinel
lymph nodes in breast cancer. Acta Cytol 1997; 41:
1606–1607 (Abstract).
58 Diamond GA. Limited assurances. Am J Cardiol 1989; 63:
99–100.
59 Lijmer JG, Mol BW, Heisterkamp S, Bonsel GJ, Prins MH,
van der Meulen JH et al. Empirical evidence of
design-related bias in studies of diagnostic tests. JAMA 1999;
282: 1061–1066.
60 Bochner MA, Kollias J, Gill PG, Farshid G, Dodd TJ.
Implication of intraoperative sentinel node imprint cytology
for consent in the SNAC Trial. ANZ J Surg 2004; 74:
105–107.
61 Chu KU, Turner RR, Hansen NM, Brennan MB, Bilchik A,
Giuliano AE. Do all patients with sentinel node metastasis
from breast carcinoma need complete axillary node
dissection? Ann Surg 1999; 229: 536–541.
62 Grube BJ, Giuliano AE. Observation of the breast cancer
patient with a tumor-positive sentinel node: implications of
the ACOSOG Z0011 trial. Semin Surg Oncol 2001; 20:
230–237.
63 Krag DN, Julian TB, Harlow SP, Weaver DL, Ashikaga T,
Bryant J et al. NSABP-32: phase III, randomized trial
comparing axillary resection with sentinel lymph node
dissection: a description of the trial. Ann Surg Oncol 2004;
11(Suppl): 208S–210S.
64 Ludwig Breast Cancer Study Group. Prognostic importance
of occult axillary lymph node micrometastases from breast
cancers. Lancet 1990; 335: 1565–1568.
65 Clare SE, Sener SF, Wilkens W, Goldschmidt R, Merkel D,
Winchester DJ. Prognostic significance of occult lymph node
metastases in node-negative breast cancer. Ann Surg Oncol
1997; 4: 447–451.
66 Sterne JC, Smith GD. Sifting the evidence: what’s wrong
with significance tests? BMJ 2001; 322: 226–231.
www.bjs.co.uk British Journal of Surgery 2005; 92: 1068–1080