ANATOMIC
PATHOLOGY
Original
P r o g n o s t i c
N u c l e a r
V a l u e
Article
o f
A n t i g e n
P r o l i f e r a t i n g
I n d e x
S t r o m a l
Correlation
With
Mitotic
i n
C e l l
G a s t r i c
T u m o r s
Count
and
Clinical
Outcome
MAHUAL B. AMIN, MD, CHAN K. MA, MD, MICHAEL D. LINDEN, MD,
JAMES J. KUBUS, MS, AND RICHARD J. ZARBO, MD
Morphologic studies of gastric stromal tumors (GSTs) indicate that
mitotic counts (MCs) and tumor size are major discriminants predictive
of biologic behavior. The authors evaluated the tumor proliferation of
GSTs with anti-proliferating cell nuclear antigen (PCNA; DAKO
clone POO, DAKO Corporation, Carpinteria, CA) for correlation with
MCs, histologic cell type, and clinical outcome. Fifty-eight tumors
ranging from 1.5 to 45 cm in size were selected for clinicopathologic
assessment. Mitotic activity was counted per 50 high-power fields
(MC). For this study, combined parameters of MC and tumor size were
used to categorize tumors into three groups: (1) benign: MC less than 5,
tumor smaller than 5 cm; (2) borderline: MC less than 5, tumor larger
than 5 cm; and (3) malignant: MC greater than 5, tumor any size. The
PCNA tumor proliferation index (TPI) was assessed from evaluation of
200 tumor cells per case and expressed as the percentage of cells with
positive results. Clinical follow-up was available in 45 cases. None of
the 19 benign or 16 borderline tumors recurred or metastasized,
whereas 7 of 10 malignant tumors metastasized and 1 of 10 recurred.
The mean PCNA TPI values among benign (11.2%), borderline (16%),
and malignant (34.5%) tumors were significantly different (P = 0.0002,
Kruskal-Wallis test). When the pathologic tumor categories were compared, the mean TPI of benign tumors was significantly different from
that of borderline tumors (P = 0.0306, Kruskal-Wallis), and the TPI of
Predicting the biologic behavior of gastric stromal tumors
(GSTs) on the basis of morphologic observations alone can be
difficult. Results of several large studies of GSTs in the English
language literature indicate that mitoses, tumor size, cellularity, and necrosis are important prognostic parameters.1"7 Despite these observations, there are rare, morphologically benign
tumors2"4,8,9 that have distant metastasis or recurrence. The
need for a third category of borderline tumors or tumors of
From the Department of Pathology, Henry Ford Hospital,
Michigan.
Presented in part at the United States and Canadian Academy of
Pathology Annual Meeting, Atlanta, Georgia, March, 1992.
Received April 24, 1992; revised manuscript accepted for publication December 21,1992.
Address reprint requests to Dr. Ma: Department of Pathology,
Henry Ford Hospital, 2799 W. Grand Boulevard, Detroit, MI 48202.
borderline tumors was different from that of the malignant tumors (P =
0.0060, Kruskal-Wallis test). The Spearman rank correlation showed a
significant relationship between the MC and PCNA TPI (P = 0.0003, r
= 0.4543). Logistic regression analysis showed that the TPI, independent of MC and size, contributed significantly (P = 0.00295) to the
prediction of outcome. In the malignant group, the mean TPI for malignant tumors with metastases (43.6%) was significantly different-^/"—
0.0411, Kruskal-Wallis test) from that of malignant tumors without
metastases (including the case with probable recurrence) (11.83%). No
correlation was found when PCNA TPIs for epithelioid GCTs were
compared with those of spindle cell GSTs. The authors conclude the
following: (1) the PCNA-derived TPI correlates with the MC and is an
additional independent prognostic parameter in the assessment of
GSTs; (2) higher PCNA TPIs in histologically malignant GSTs may
correlate with metastasis and clinical outcome; (3) in GSTs with an MC
less than 5, the PCNA TPI provides a quantitative parameter to potentially separate borderline from benign tumors; and (4) the intermediate
PCNA TPI of borderline tumors defined by size further justifies separation of GSTs into benign, borderline, and malignant categories. (Key
words: Gastric stromal tumors; Cell proliferation; Proliferating cell nuclear antigen) Am J Clin Pathol 1993;100:428-432.
unknown malignant potential has been recognized because of
difficulty in predicting outcome based on histologic findings
alone.
In recent years, measurements of tumor proliferation have
shown a good correlation with biologic behavior in certain
solid tumors, providing valuable adjunctive prognostic information.10'11 In this study, we measured tumor proliferation
within GSTs by assessing immunostaining of tissue sections
fixed in formaldehyde solution and embedded in paraffin with
anti-proliferating cell nuclear antigen (PCNA; DAKO clone
PC 10, DAKO Corporation, Carpinteria, CA) and correlating
Detroit,
this tumor proliferation measurement with the MC and clinical outcome. Proliferating cell nuclear antigen is a cell cycle-related nuclear protein identified as an auxiliary protein of DNA
polymerase delta that is maximally elevated in late Gl- and
S-phases of cycling cells.12
Based on a literature review of GSTs, the most significant
pathologic discriminants of biologic behavior include
MC1,2,13,14 a n d t u m o r s J ze 2,6,14-16 W g use£ j t h e s e c r j t e r j a tO di-
428
AMIN ET AL.
429
PCNA Index in GSTs
vide GSTs into benign, borderline, and malignant categories.
Tumors also were classified, based on the predominant tumor
cell type, as having either epithelioid or spindle cell histologic
characteristics. The objectives of this study were as follows: (1)
to measure PCNA tumor proliferation indices (TPIs) for benign, borderline, and malignant categories; (2) to correlate the
PCNA TPI of tumors with the MC, tumor size, and spindled
and epithelioid histologic characteristics; and (3) to determine
whether the PCNA TPIs of GSTs are correlated with clinical
outcome and provide additional prognostic information.
MATERIALS AND METHODS
Seventy cases of GST were retrieved from the Surgical Pathology files of Henry Ford Hospital, Detroit, Michigan, from
1966 to 1991. Twelve GSTs were found incidentally during
surgery for other diseases and were smaller than 1 cm (most
were a few millimeters). They had no clinical significance and,
therefore, were excluded from our study. Fifty-eight cases of
GST remained, with sizes ranging from 1.5 to 45 cm. These
cases form the basis of this clinicopathologic and immunohistologic study.
Histologic Examination and Clinical Data Review
Each tumor was examined histologically and assessed for
cellularity, degree of anaplasia, pleomorphism, necrosis, MC,
and predominant cell type (epithelioid or spindled). Areas with
the greatest mitotic activity were selected, and MCs were established from 100 consecutive high-power fields (HPFs) (Olympus BH-2; X40 high dry, X10 ocular; 0.2-mm2fieldsize, Olympus Corp., Lake Success, NY). The MC thus obtained was
divided by two and denoted the number per 50 HPFs. Based on
our review of the literature, we determined that MC1,213,14 and
tumor size2'6'4"16 were the two most important prognostic parameters. We used Appelman's determination of 5 mitoses per
50 HPFs to separate benign from malignant lesions.1,2 Tumor
size also is predictive of the clinical behavior of GSTs. Thus, we
categorized GSTs as follows: benign: less than 5 mitoses/50
HPFs and tumor smaller than 5 cm; borderline: less than 5
mitoses/50 HPFs and tumor larger than 5 cm; and malignant:
more than 5 mitoses/50 HPFs and tumor of any size. The clinical charts and Tumor Registry were reviewed to obtain information on clinical outcome.
Itnmunohistochemical
Staining
In each case, one representative block was chosen for immunohistochemical staining with the Streptavidin method
(DAKO). Tissue sections were mounted on gelatin-coated
slides, air dried, deparaffinized in xylene, and rehydrated with
distilled water and decreasing concentrations of ethanol. The
slides were air dried because oven heating was found to reduce
optimal staining. Enzymatic predigestion was not performed
because it also was found to diminish the staining intensity.
The slides then were incubated with 5% horse serum (suppressor serum) and incubated for 14-16 hours overnight at 4 °C
with a 1:1,500 dilution of monoclonal antibody to PCNA
(DAKO clone PC 10). Sections then were incubated for 30 minutes with biotinylated horse antimouse IgG at room temperature in a dark, humid, closed chamber and then for 60 minutes
at room temperature with Streptavidin. Sections were developed for five minutes with the chromogen 3-amino-9-ethyl-
carbazole (Sigma Chemical Company, St. Louis, MO), lightly
counterstained with Mayer's hematoxylin, and coverslipped
with glycerol gelatin. A lymph node with reactive follicular
hyperplasia served as a positive control; in negative controls,
buffer was substituted for primary antibody.
For each of the 58 tumors, a proliferation index indicating
the percentage of positive tumor nuclei was obtained by visual
quantitative assessment of 100 tumor cells by two observers
(M.B.A. and M.D.L.) on separate occasions, and the mean of
the two values was designated as the TPI. Random fields were
selected, excluding areas with excessive or scant immunostaining. Care was taken to exclude positively stained inflammatory
cells and desmoplastic stromal cells; in problem cases, this was
achieved by comparison with the corresponding hematoxylin
and eosin-stained sections. Nuclear staining without cytoplasmic staining was considered a positive result. Staining intensity
varied from very weak to intensely strong and fromfinelygranular to uniformly dark red-brown.
Statistical
Analysis
Statistical tests were performed with CSS (Stat Soft, Tulsa,
OK) software on a microcomputer. A Kruskal-Wallis test was
performed to determine significant differences or associations
of TPI among the following: (1) benign, borderline, and malignant categories together and individually with respect to each
other; (2) malignant tumors with and without metastasis; and
(3) tumors with spindled and epithelioid histologic characteristics.17 A Spearman rank correlation was performed to determine whether MC and TPI were correlated. Logistic regression
analysis was used to assess whether the TPI contributed significantly to the prediction of outcome (ie, recurrence of disease or
metastasis) independently of size and MC.
RESULTS
Based on a detailed histologic review and analysis with criteria outlined above, there were 24 benign, 18 borderline, and 16
malignant tumors. All benign tumors were diagnosed originally as benign. Of the 18 borderline tumors, 11 were interpreted as benign, 3 as malignant, and 4 as suggestive of malignancy. Among the 16 malignant tumors, 15 were diagnosed as
malignant and 1 as atypical leiomyoblastoma (this patient died
one year after surgery with probable metastasis to the lungs).
Generally, the malignant tumors were characterized by a significant increase in cellularity, pleomorphism, and cellular atypia.
Compared with benign tumors, the borderline tumors tended
to be more cellular and had cellular atypia; however, these
parameters overlapped between each category. The tumor sizes
and MCs of tumors within benign, borderline, and malignant
categories are summarized in Table 1.
Clinical data were available for 45 cases with follow-up of
1-24 years. None of the 19 benign or 16 borderline tumors
recurred or metastasized. One borderline tumor had infiltrated
the liver at the time of surgery and required wider excision with
partial lobectomy. This tumor had not metastasized or recurred four years after surgery, after which the patient was lost
to follow-up. Of the patients with malignant tumors, seven of
ten had distant metastasis, two of ten had no recurrence or
metastasis, and one often had probable clinical recurrence (not
histologically confirmed).
Overall, immunohistochemical staining patterns (Fig. 1)
were variable and heterogeneous from field tofieldand case to
Vol. 100-No. 4
430
ANATOMIC PATHOLOGY
Original Article
TABLE 1. SIZE, MITOTIC COUNT, AND PCNA TPI OF GST
Size (cm)
Benign (n = 24)
Borderline (n = 18)
Malignant (n = 16)
PCNA TPI (%)
Mitotic Count*
(MC)
Range
Mean
Median
1.5-4.9
5-33
3-45
3.26
12.33
15.6
3.5
11
12.5
Range
0-2.5
0-4.5
5-54
Mean
0.63
1.13
22.2
Median
0.25
0.75
21.3
Range
Mean
0-69.5
1.5-47.5
1.5-84
11.2
16
34.5
Median
3.3
12
29.8
PCNA = proliferation cell nuclear antigen; TPI = tumor proliferation index; GST = gastric stromal tumor.
* Mitotic count in 100 HPFs was determined and an MC (mitotic count/50 HPF) was obtained.
case; within and among benign, borderline, and malignant categories; and between histologic cell types (spindle and epithelioid). The mean TPIs for various categories varied considerably:
benign (range, 0-69.5%; mean, 11.2%); borderline (range, 1.547.5%; mean, 16%); and malignant (range, 1.5-84%; mean,
34.5%) (Fig. 2). The Kruskal-Wallis statistical analysis showed
that the three categories were significantly different (P =
0.0002) and there was a significant difference between benign
and borderline categories (P = 0.0306) and borderline and malignant categories (P = 0.0060).
The Spearman rank correlation test showed a significant
correlation between MC and TPI (P = 0.0003; r = 0.4543). The
mean TPI of malignant tumors with metastasis (43.6%) was
significantly different from the mean TPI of malignant tumors
without metastasis (11.83%) (P = 0.0411, Kruskal-Wallis test)
(Fig. 3). The mean TPIs for tumors with spindled and epithelioid histologic characteristics were not significantly different (P
= 0.5895, Kruskal-Wallis test). The logistic regression model
showed that MC, size, and TPI all predicted outcome significantly (P = 0.00002). It also showed that TPI alone contributed
significantly (P = 0.00295) to the prediction of outcome more
than the model that included only MC and size. The sensitivity
and specificity of the logistic model were 75% and 95%, respectively, in predicting outcome. Of the 45 cases with data regarding history and histologic characteristics, the model predicted
that 37 would show no evidence of disease during the follow-up
and 8 would metastasize or recur. Of the 37 patients predicted to have no evidence of disease, 2 actually had metastasis,
and of the 8 predicted to have metastasis or recurrence, only 6
actually did.
DISCUSSION
Gastric stromal tumors are enigmatic in terms of their line of
differentiation or cell of origin and clinical behavior.13 The
histologic distinction of benign from malignant GSTs is often
difficult, and prediction of clinical outcome on the basis of
histologic characteristics alone is not always reliable. Therefore, a third category for borderline tumors or tumors of unknown malignant potential has been suggested in the pathologic diagnosis of GSTs. Many studies have shown histologic
features such as cellularity, necrosis, tumor size, and mitoses to
be prognostically significant parameters.1"7 Pathologic criteria,
including the number of mitoses/HPF, have been proposed to
differentiate benign from malignant tumors. Mitotic count criteria are not applied universally and often vary from series to
series. Although the MC appears to be the single most reliable
criterion of GST biologic behavior,5-6,13'4 the ability to predict
outcome is not absolute.2~4-8-9 There is compelling evidence indicating that tumor size is another important prognostic deter-
FIG. 1. Malignant GST, showing many
positively stained tumor cells. A. Spindle type. Proliferating cell nuclear antigen (original magnification, X80). B.
Epithelioid type. Proliferating cell nuclear antigen (original magnification,
X80).
A.J.C.P.-October 1993
431
AMIN ET AL.
PCNA Index in GSTs
70
60
JL.
50
40
30
20
O-
10
BENIGN
BORDERLINE
MALIGNANT
FlG. 2. Ranges and means of TPIs for benign, borderline, and malignant GSTS.
minant.2-614"16 Based on our review of the literature, we divided
GSTs into three categories—benign, malignant, and borderline
—using MC and tumor size as the best determinants of biologic behavior. Assessment of PCNA immunostaining was
used to determine whether this measurement of tumor proliferation was of prognostic value and correlated with MC and tumor size.
Enumeration of mitotic figures has been the conventional
means of assessing cellular proliferation in many neoplasms.
Unfortunately, mitotic counting is fraught with inaccuracy and
lacks reproducibility. Mitotic counts also may be affected by
tissue fixation, microscope field size, tumor heterogeneity,18'9
and cutoff values segregating malignant tumors from benign.
Although, as a measure of cellular proliferation, mitosis is
currently the "histologic gold standard," attention is being focused on other methods to determine tumor proliferation.
Row cytometric analysis, computerized static image analysis,
the tritiated thymidine labeling index, and silver-stained nucleolar organizer regions are some of the techniques currently
under active investigation.1018 Immunohistochemical identification of nuclear antigens associated with cellular proliferation
offers an alternative approach for measuring the tumor growth
fraction. For routine pathology practice, the monoclonal antibody Ki-67, an incompletely characterized nuclear protein associated with cell proliferation, is currently the best immunohistochemical method and has been shown to correlate well with
other measures of cell proliferation20; however, the Ki-67 antibody requires fresh tissue and cannot be used in tissue sections
fixed in formaldehyde solution and embedded in paraffin.
Recently PCNA (cyclin), a cell cycle-associated 36-kD nuclear protein identified as an auxiliary protein of DNA polymerase delta, has been shown to be an effective indicator of cell
proliferation and can be identified by immunohistochemical
staining in tissue sections fixed in formaldehyde solution and
embedded in paraffin."13'2'"25 PCNA levels increase through
G,, peak at the G,/S-phase interface, decrease through G2, and
reach virtually undetectable levels by immunocytochemical
methods in M-phase and quiescent cells.12'23'24 Proliferating
cell nuclear antigen levels have been shown to correlate with
90 n
•
i
©
mban
i
80-
|
70
60*n
|
|
|
|
|
|
|
|
!
(j>
0
....
80
mitotic activity, tumor grade,13,26"28 and various growth fractions as measured by flow cytometric analysis.22-27 Initial clinical studies in gastric carcinomas,26 gastric lymphomas,27 gastrointestinal stromal tumors,29 and hemangiopericytomas28
suggest a role for PCNA staining as an adjunctive prognostic
indicator.
In this study, the mean TPIs of benign, borderline, and malignant tumors were significantly different. The pathologic difference between benign and borderline tumors was based on the
size of the tumors alone (all had MCs < 5) with a cutoff of 5 cm.
The intermediate TPIs of borderline GSTs, differing only from
the benign tumors because they were larger, seem to further
justify the separation of GSTs into three categories. The statistically significant difference in mean TPIs between these two
tumor categories suggests that tumor size may be an important
prognostic determinant in GSTs. In tumors of any size with
MCs less than 5, the PCNA TPI appears to provide a quantitative parameter, in addition to tumor size differences, for separating benign from borderline GSTs. This was not possible with
MCs alone because the benign and borderline tumors both had
MCs less than 5; however, the duration of follow-up in some of
the cases in benign and borderline categories was short, possibly compromising our ability to show the value of PCNA immunostaining in predicting differences in the outcome of tumors with similar MC ranges of less than 5 per 50 HPFs. The
mean TPI of tumors with spindled histologic characteristics
was not significantly different from that of GSTs with epithelioid histologic characteristics, suggesting that the two histologic
cell types are merely different morphologic expressions of
GSTs without biologic significance.
The mean TPI of malignant tumors with metastases was
significantly different than the TPI of malignant tumors without metastases (P = 0.0303); thus, the mean TPI may be a
potential predictor of metastatic potential. The number of
cases studied, however, is too small to allow us to draw any firm
conclusions. This observation regarding the possible prognostic value of PCNA is similar to findings from the study by Yu
and colleagues28 of hemangiopericytomas, where PCNA staining identified patients at risk of metastasis and early death. A
TPI
90-
I
8i
!
20
d>
i
|
10
•
i
30 "
0-
• • • • ! • •
. . f . . .
W/O METASTASIS
i
.
i
i
i
i -1-.
i
i
i -i
i
i
W/ METASTASIS
FIG. 3. Ranges and means of TPIs for malignant GSTs with and without metastasis.
Vol. 100-No. 4
432
ANATOMIC PATHOLOGY
Original Article
significant correlation was found between the MC and PCNA
TPI, an observation consistent with other studies of solid tumors using PCNA,12 thus suggesting PCNA TPI as an additional prognostic parameter for these enigmatic tumors. The
logistic regression analysis suggests that the TPI, independent
of the MC and size, contributes significantly to the prediction
of outcome. The regression model seems more useful in predicting no evidence of disease than for metastasis or recurrence, as indicated by the relatively high specificity. Additional
studies must be performed to determine the relative value of
size, MC, and TPI in predicting outcome, especially survival.
Despite the fact that the aggregate evidence indicates some
biologic and clinical utility of PCNA staining, its application
for an individual case with currently available technology is
limited. We observed significant variation of PCNA TPIs for
individual cases and for the benign, borderline, and malignant
tumor categories. Multiple factors may be responsible for this
variation, including length and type of fixation, sampling bias,
intratumoral heterogeneity, and visual thresholds in the assessment of positive staining. Variable staining intensity within a
tumor also may result from differences in antigen preservation.
Previous studies have found that PCNA staining may vary with
length of fixation and preparatory techniques.25 Studies with a
larger number of cases—as well as standardization of preparatory techniques and methods of quantitation of TPI, and reduction in interlaboratory and intralaboratory variation of results30—must be performed before PCNA immunostaining
can be used as a reliable adjunct in routine surgical pathology
practice.
11. Hall PA, Levison DA. Assessment of cell proliferation in histological material. J Clin Pathol 1990;43:184-192.
12. Robbins BA, Vega D, Ogata K, et al. Immunohistochemical detection of proliferating cell nuclear antigen in solid human malignancies. Arch Pathol Lab Med 1987; 111:841-845.
13. Antonioli DA. Gastrointestinal autonomic nerve tumors: Expanding the spectrum of gastrointestinal tumors itorial]. Arch Pathol
Lab Med 1989;113:831-833.
14. Shiu MH, FarrGH, Papachristou BN, Hajdu SI. Myosarcomas of
the stomach: Natural history, prognostic factors and management. Cancer 1982;49:177-187.
15. Haque S, Dean P. Stromal neoplasms of the rectum and anus. Lab
Invest 1989;60:38A(Abstr).
16. Federspiel BH, Sobin LH, Helwig EB, et al. Morphometry and
cytophotometric assessment of DNA in smooth muscle tumors
(leiomyomas and leiomyosarcomas) of the gastrointestinal
tract. Anal Quant Cytol Histol 1987;9:105-114.
17. Sokal RR, Rohlf FJ. Biometry. San Francisco: WH Freeman and
Company, 1981, pp 429-432.
18. Woosley JT. Measuring cell proliferation. Arch Pathol Lab Med
1991;115:555-557.
19. Silverberg SG. Reproducibility of the mitosis count in the histologic diagnosis of smooth muscle tumors of the uterus. Hum
Pathol 1976;7:451-454.
20. Hitchcock CL. K.i-67 staining as a means to simplify analysis of
tumor cell proliferation itorial]. Am J Clin Pathol 1991;96:444445.
21. Celis JE, Bravo R, Larsen PM, Fey SJ. Cyclin: A nuclear protein
whose level correlates directly with the proliferative state or normal as well as transformed cells. Leuk Res 1984;8:143-157.
22. Garcia RL, Coltrera MD, Gown AM. Analysis of proliferative
grade using anti PCNA/cyclin monoclonal antibodies in fixed,
embedded tissues: Comparison with flow cytometric analysis.
Am J Pathol 1989; 134:733-739.
Acknowledgments. The authors thank Kathy Zajas for her secretarial 23. Waseem NH, Lane DP. Monoclonal antibody analysis of the prosupport and Nancy Peshkin, Kathy Roszka, and Monica Grenke for
liferating cell nuclear antigen (PCNA): Structural conservation
technical assistance.
and the detection of a nucleolar form. J Cell Sci 1990;96:121129.
24. Coltrera MD, Gown AM. PCNA/cyclin expression in BrdU upREFERENCES
take defines different subpopulations in different cell lines. J
Histochem Cytochem 1991;39:23-30.
1. Appelman HD, Helwig EB. Sarcomas of the stomach. Am J Clin
25. Hall PA, Levison DA, Woods AL, et al. Proliferating cell nuclear
Pathol 1977;67:2-10.
antigen (PCNA) immunolocalization on paraffin sections: An
2. Appelman HD, Helwig EB. Gastric epithelioid leiomyoma and
index of cell proliferation with evidence of deregulated expresleiomyosarcoma (leiomyoblastoma). Cancer 1976;38:708-728.
sion in some neoplasms. J Pathol 1990; 162:285-294.
3. Appelman H D, Helwig EB. Cellular leiomyomas of the stomach in 26. Jain S, Filipe MI, Hall PA, et al. Prognostic value of proliferating
49 patients. Arch Pathol Lab Med 1977; 101:373-377.
cell nuclear antigen in gastric carcinoma. J Clin Pathol
4. Morgan BK, Compton CC, Talbert M, et al. Benign smooth mus1991;44:655-659.
cle tumors of the gastrointestinal tract: A 24-year experience.
27. Woods AL, Hall PA, Shepherd NA, et al. The assessment of proAnn Surg 1990;211:63-66.
liferating cell nuclear antigen (PCNA) immunostaining in pri5. Evans HL. Smooth muscle tumors of the gastrointestinal tract: A
mary gastrointestinal lymphomas and its relationship to histostudy of 56 cases followed for a minimum of 10 years. Cancer
logical grade, S+G2+M phase fraction (flow cytometric
1985;56:2242-2250.
analysis) and prognosis. Histopathology 1991;19:21-27.
6. McGrath PC, Neifeld JP, Lawrence W, et al. Gastrointestinal sar28. Yu CCW, Hall PA, Fletcher CDM, et al. Hemangiopericytomas:
comas: Analysis of prognostic factors. Ann Surg 1987;206:706The prognostic value of immunohistochemical staining with a
710.
monoclonal antibody to proliferating cell nuclear antigen
7. Ranchod M, Kempson RL. Smooth muscle tumors of the gastroin(PCNA). Histopathology 1991;19:29-33.
testinal tract and retroperitoneum. Cancer 1977;39:255-262.
29. Yu CCW, Fletcher CDM, Newman PL, et al. A comparison of
8. Schweitzer G. Smooth muscle tumors of the alimentary tract. NZ J
proliferating cell nuclear antigen (PCNA) immunostaining, nuSurg 1967;36:218-222.
cleolar organizer region (AgNOR) staining, and histological
9. Miller KA, Rubnitz ME, Roth SI. Late recurrence (33 years) of a
grading in gastrointestinal stromal tumours. J Pathol
gastric epithelioid stromal tumor (leiomyoblastoma) with low
1992;166:147-152.
malignant potential. Arch Pathol Lab Med 1988; 112:86-90.
30. Linden MD, Torres FX, Kubus J, Zarbo RJ. Clinical application
10. Meyer JS. Cell kinetic measurements of human tumors. Hum
of morphologic and immunocytochemical assessments of cell
Pathol 1982;13:874-877.
proliferation. Am J Clin Pathol 1992;97 (suppl 1): S4-S13.
A.J.C.P. • October 1993