Jpn J Clin Oncol 2002;32(9)347–351

Helpful Parameter for Malignant Potential of Gastrointestinal Stromal Tumors (GIST) Xiaojuan Wang1, Ichiro Mori1, Weihua Tang1, Hirotoshi Utsunomiya1, Misa Nakamura1, Yasushi Nakamura1, Genying Zhou2 and Kakudo Kennichi1 1Department

of Pathology, Wakayama Medical University, Wakayama City, Japan and 2Department of Pathology, School of Medicine, Shandong University, Jinan, China Received October 15, 2001; accepted May 30, 2002

Background: Although a series of histopathological criteria have been suggested, the prediction of the malignant potential of gastrointestinal stromal tumors (GIST) is still difficult. The older literature called all gastrointestinal stromal tumors smooth muscle tumors or mixed GIST with true smooth muscle tumors. Reports on GIST including homogeneous cases were rare. Methods: We examined 73 cases of GIST, which were immunohistochemically positive for c-kit and/or CD34, and mainly focused on the correlation between mitotic count and the other clinicopathological features to establish any helpful and reproducible parameters to indicate the malignant potential and to be used practically and objectively in the routine histopathological diagnosis of GIST. Results: The results showed that there was a statistically significant difference in mitotic count between benign and malignant groups. Other proposed parameters, such as high cellularity, tumor size ³5 cm, stomach and intestinal location, hemorrhage, necrosis, p53 expression and Ki-67 labeling index >10%, were frequently observed in tumors with mitotic figure. Three patients with one mitotic figure in 50HPF died from metastasis or recurrence of the tumors. Conclusions: Ki-67 index and cellularity should be used as predictors for the malignant potential of GIST. When other morphological features appear benign, mitotic count might also be a helpful practical factor in the prediction of the malignant potential of GIST.

Key words: gastrointestinal stromal tumors – alimentary tract – mitotic count – prognosis – malignant potential

INTRODUCTION Gastrointestinal stromal tumors (GIST) are the most common non-epithelial type of tumor among those arising from the gastrointestinal tract. The histological origin of this tumor has recently been suggested to be the interstitial cells of Cajal (1) because of its expression of c-kit and/or CD34. A series of histopathological criteria have been reported to predict the malignant potential of these tumors but their conclusions varied. It was reported that the prognosis of the GIST correlated well with mitotic count, tumor size, tumor cellularity, tumor necrosis, anatomical location, invasive growth (2–7) and expression of Ki-67 (8–12) and PCNA index (13). However, diagnosis and prediction of the malignant potential of GIST is still difficult. On the other hand, the older literature called all stromal

For reprints and all correspondence: Xiaojuan Wang, Department of Pathology, Wakayama Medical University, 811–1 Kimiidera, Wakayama City 641-0012, Japan. E-mail: [email protected]

tumors smooth muscle tumors or mixed GIST with true smooth muscle tumors, because the latter often occur in the esophagus and show clearer biological behavior. The study of homogeneous groups of true GIST has therefore been advocated (14). The aim of this study was to establish any helpful and reproducible parameters to indicate malignant potential and to be used practically and objectively in the histopathological diagnosis of GIST. We examined 73 cases of GIST immunohistochemically positive for c-kit and/or CD34 and mainly focused on the correlation between mitotic count and other proposed clinicopathological features. Those cases negative for c-kit or CD34 were excluded from the study.

PATIENTS AND METHODS Seventy-three consecutive cases of GIST were selected from 106 cases of tumors previously diagnosed as leiomyoma, leiomyoblastoma and leiomyosarcoma from the files of the Department of Pathology, Shandong University, China, in the 30-year period from 1969 to 1999, based on immunohisto© 2002 Foundation for Promotion of Cancer Research

348

Prognosis for GIST

Table 1. Antibodies, dilutions and sources in the present study Antibody

Clone

Pretreatment

Dilution

Source*

c-kit

Poly

Microwave

1:20

1

CD34

Qbend10

Microwave

1:25

2

SMA

1A4

None

1:400

3

S-100

Poly

None

1:500

4

Ki-67

MIB-1

None

1:50

5

P53

DO7

Microwave

1:400

4

*1, Immuno-Biochemical Laboratories, Fujioka, Japan; 2, Novocastra, Newcastle, UK; 3, Sigma Chemical, St Louis, MO, USA; 4, DAKO, Kyoto, Japan; 5, Immunotech, Marseilles, France.

chemical studies on c-kit and CD34. All specimens were from patients who underwent completely surgical excision without record of positive margin. All of these surgical specimens were routinely fixed in formalin and embedded in paraffin. At least 4–5 paraffin blocks were available in each case for study. Clinical information including tumor size was obtained from surgical and pathological records. Tumors that developed recurrence or metastasis were judged as malignant, including those which induced death of the patient. Tumors with peripheral invasive growth microscopically were also diagnosed as malignant. The other cases without the above evidence of malignancy were classified as benign in this study, although there was still a slight chance of malignancy in the benign group. No-mitosis group, low-mitosis group and high-mitosis group were defined depending on mitotic numbers as described previously (4,5,13,14). Briefly, 10 high-power fields in one group were observed to find the mitotic figure at high cellularity fields; a total of five groups (50 high-power fields) were observed in each case. No-mitosis cases virtually showed no mitosis, i.e. 0/50HPF, low-mitosis cases showed 1–4 mitotic figures in 50HPF and high-mitosis cases showed >5 mitotic figures in 50HPF. Simultaneously, (1) histological pattern: leiomyoma-like type, Schwannoma-like type, leiomyoblastomalike type and mixed-type; (2) cellularity: low (I), intermediate (II) and high (III); (3) tumor necrosis: presence or absence; and (4) tumor hemorrhage: presence or absence, were also investigated in each case. A standard labeled streptavidin–biotin (LSAB) method of immunohistochemistry was used. The dilution, sources and pretreatment of antibodies are summarized in Table 1. Positive and negative controls were used properly. For p53, the results were classified into three groups (no staining for negative, distinctly stained cells 50% for two plus positive). For Ki-67 index, the results were classified into two groups (distinctly stained cells 10% for another group). Time-independent continuous variables were evaluated using Student’s t-test and the chi-squared test was used for categorical data. The F statistic was tested to estimate the cor-

Figure 1. (a) No-mitotic group. The tumor is mainly composed of spindle cells, with low cellularity and few mitotic figures. (b) Low-mitotic group. The tumor is composed of spindle cells and short rod cells, with moderate cellularity. Mitotic figures are seen occasionally (1–4/10HPF). (c) High-mitosis group. The tumor is mainly composed of short rod cells and irregular cells with high cellularity. Mitotic figure is easier to find (>5/10HPF).

relation between mitosis and the other parameters. Survival analysis was done with the StatView-J 5.0 statistical software package (SAS Institute, Cary, NC). For univariate analysis of time-dependent variables, the Kaplan–Meier method and the log-rank test were used to determine significant prognostic factors. In multivariate analysis, a Cox proportional hazards

Jpn J Clin Oncol 2002;32(9)

349

Table 2. Clinical, histochemical, immunohistochemical and follow-up results in the study Benign (n = 38)

Malignant (n = 35)

P value

Age (years)

48.9

51.8

NS

Male:female

19:19

23:12

NS

Mean tumor size (cm)

5.31

9.20

0.0001

Location (E:S:SI:LI)

12:16:7:3

1:18:15:1

E vs others, P = 0.0015

Histological pattern (Lei:others)

30:8

23:12

NS

Cellularity (low:mid + high)

21:17

9:26

0.0168

No-mitosis (0/10HPF)

24/38

0/35

5/10HPF)

2/38

17/35