Impact of serosal involvement/extramural growth on the risk of synchronous and metachronous peritoneal spread in gastrointestinal stromal tumors: proposal for a macroscopic classification of GIST

Abstract

Background

Peritoneal and hepatic metastases are the main routes of spread of gastrointestinal stromal tumors (GIST). However, criteria to predict the site and pattern of recurrence in individual cases are still lacking.

Patients

We retrospectively analyzed 67 consecutive GISTs with complete gross descriptions to correlate macroscopic patterns with clinical course. Primary endpoint was the appearance of synchronous or metachronous peritoneal disease. Tumors were classified into type I (luminal/intramural) and type II (extramural) based on the macroscopic/histologic presence or absence of normal tissue between deeper tumor border and serosa, respectively.

Results

Patients were 35 men and 32 women (mean age, 64 yrs) with gastric (n=32), small bowel (n=30) and large bowel (n=5) GISTs. Based on the above proposal, 22 tumors were classified as type I and 45 as type II. Type I tumors were predominantly gastric (18/22; P<0.001) and frequently had very low/low risk (14/22; P<0.001) whereas type II tumors were predominantly intestinal (31/45; P<0.001) and often of intermediate/high risk (36/45; P<0.001). Ten patients had synchronous peritoneal spread and 6/30 patients with a mean follow-up of 29 months developed metachronous peritoneal spread at a mean of 27 months. Tumor rupture was seen in 2 patients (3%). Thus, 16/40 patients (40%) had synchronous or metachronous peritoneal progression. Taken by gross type, peritoneal progression was seen in 15/30 type II compared to 1/10 type I tumors (p=0.032).

Conclusion

this study points to extramural growth as a predictor of peritoneal recurrence in GIST, probably as a consequence of tumor rupture or due to microscopic serosal penetration. This study aimed at alerting surgical pathologists to the importance of careful gross and microscopic assessment of resection specimen harboring GIST to allow for reliable prospective evaluation of serosal involvement as an adverse prognostic factor in GIST.

Introduction

Although relatively rare, gastrointestinal stromal tumors (GISTs) represent the most common primary mesenchymal neoplasms of the tubular GI tract [1]. Their estimated annual incidence ranged from 14 to 20 new cases per 1 million per year [2]. Approximately one third of GISTs follow a malignant behavior during the clinical course of the disease; half of them show evidence of metastasis at the time of diagnosis [3]. Assessment of the malignant potential of GIST still represents a major challenge to pathologists, surgeons and oncologists involved in the management of GIST patients. Accordingly, several schemes have been proposed by different authors for risk stratification in a trial to reliably identify those patients at a higher risk for disease relapse and who would hence benefit from appropriate surgical and/or adjuvant treatment [1, 4-9]. Several clinicopathological parameters have been used to predict disease recurrence; the most important of them are tumor size, anatomic site and mitotic count in 50 high power fields (HPFs) [1]. In addition, several other facultative parameters have been shown to be associated with adverse outcome by some authors such as MiB-1 index [10], histological subtype [11], alterations related to specific on-coproteins such as p16 [12], unfavorable kinase mutations [13] and cytogenetic aberrations [14]. Of the above risk systems, the classification proposed by Miettinen and Lasota proved most useful in risk prediction [1]. However, while the current risk classification systems are of great value in identifying very low/low risk (mostly benign) and intermediate/high risk (mostly malignant) tumors with satisfactory reliability [15, 16], there still exists a gray zone that encompasses tumors from the intermediate category and a small fraction of tumors from the low risk category that occasionally behave in a malignant fashion despite their bland looking histology, low mitotic activity and relatively small size, thus making the above listed criteria of limited practical value in their identification [17]. Furthermore, these risk systems have not worked well in the assessment of the biological behavior in pediatric GISTs [18].

Peritoneal dissemination and hematogenous spread to the liver and, rarely, to other distant organs have been identified as the major routes of GIST spread [1, 3, 19]. Some recent studies have demonstrated that peritoneal contamination as a consequence of tumor rupture [20] or serosal penetration [21] represents an adverse prognostic factor in GISTs. However, tumor rupture was reported to affect no more than 5-7% of the patients in previous series [20]. In contrast with this, peritoneal dissemination occurred alone or in combination with liver metastasis in 61% of patients in previous series [19]. Yet, the mechanisms responsible for this relatively high rate of peritoneal dissemination at the time of initial diagnosis or during the later course of the disease in patients without overt spontaneous or intra-operative tumor rupture remain largely unknown.

In this retrospective analysis, we have tried to examine the impact of the gross pattern (as determined by the presence of luminal/intramural vs. extramural growth) on the occurrence of synchronous and metachronous peritoneal tumor spread in a group of GIST with complete gross documentation of their growth patterns.

Materials and methods

Patients' characteristics

Tumors were recovered from a retrospective GIST database encompassing tumors from the Pathology departments of Nuremberg Clinic center and the Erlangen University Hospital, Germany. Pathology reports have been thoroughly evaluated for detailed gross descriptions of tumors. Only cases with complete gross descriptions of their luminal and serosal surfaces and with available gross photographs have been included in this study irrespective of their site, size, age, gender and presence or absence of peritoneal or distant metastasis. Minute incidental gastric GISTs (<1 cm) have been excluded as these lesions are very common and carry no clinical significance [22]. Assessment of the macroscopic pattern was performed without knowledge of the disease outcome. The occurrence of peritoneal implants/metastasis distinct from the primary tumor mass (including both synchronous and metachronous peritoneal deposits/metastasis) and the development of intra-abdominal extra-gastrointestinal or retrop-eritoneal recurrence were documented as evidence of tumor progression (primary endpoint). Patients with synchronous or metachronous progressive disease and those without progressive disease at a mean follow-up of 29 months were considered to have reached the primary endpoint and were included into further statistical analysis of disease outcome (total: 40). Those who did not reach the primary endpoint or had no follow-up (n=27 patients) were included into the initial subtyping and demographic analysis but were not included into further statistical analysis of disease outcome. Risk classification has been done using the classical Fletcher et al-classification [5] as well as the classification proposed by Miettinen and Lasota [1]. Tumors with synchronous metastasis that did not undergo radical surgery were considered into the high risk group (clinically malignant tumors). Their detailed gross features were obtained from surgical reports and intra-operative gross photographs.

Description of macroscopic tumor types

For the purpose of this study and to facilitate analysis of the wide spectrum of the gross presentation of GIST, 6 possible gross types were defined to categorize patient's tumors thereby taking into consideration the diverse gross pattern reported for GIST in previously published large studies [1, 23, 24]. The six subgroups were then lumped into two main groups, each composed of three subtypes. These subgroups and their respective defining criteria are summarized in Table 1 and representative examples are depicted in Figure 1. In brief, type I tumors have in common the presence of clear-cut layer of compressed normal muscularis propria or subserosal tissue between deeper portion of the tumor tissue and the covering peritoneum/serosa. In contrast, type II tumors do not possess more than a thin serosal covering (tumor capsule or single mesothelial layer) and they have in common a variable extramural protruding component bulging into the abdominal cavity. In addition, sections taken from the peritoneal surface of the tumor have been retrospectively evaluated to confirm macroscopic findings. The number of sections varied with tumor size, but generally at least two sections were submitted from the serosal covering during tumor processing. Additional sections were submitted from areas of suspected serosal defects or rupture. For statistical purposes, tumors within the spectrum of type I have been analyzed as a group and compared to those of type II.

Table 1.

definition of the macroscopic types in GIST^*

Tumor type	Defining criteria
Type IA	Predominantly polypoid, apparently confined to the submucosa, abutting the inner muscularis propria with no more-than-minimal involvement of the muscularis propria (commonly as endoscopic specimen).
Type IB	Apparently polypoid, but with more-than-minimal involvement of the muscularis propria and/or with broad-based attachment to the latter.
Type IC	Intramural with fusiform splitting of the muscularis propria, normal tissue both luminally and serosally found.
Type IIA	Dumbbell protruding through both the mucosa and the serosa. No apparent normal tissue (compare to IC) is seen at the peritoneal border.
Type IIB	Completely extramural with a short pedicle or forming a sessile serosal mass or pseudo-diverticulum.
Type IIC	Completely intra-abdominal with unclear point of attachment to the gut wall or with subtle adhesions to a bowel segment (equivalent to the so-called extra-gastrointestinal GIST).

^*Tumors have been classified irrespective of presence or absence of tumor rupture.

Figure 1.

Representative examples of the 6 gross patterns in GISTs (numbered according to the macroscopic types defined in table 1). IA: gastric GIST with predominant polypoid (luminal) growth and minimal (microscopic) involvement of the inner muscle layer (completely removed via polypectomy; arrows point to submucosal remnants). IB: ulcerated gastric GIST with luminal/intramural growth obliterating the inner muscle layer but still separated from serosa by recognizable external muscle layer and subserosa (arrows). IC: this small bowel GIST shows a combination of luminal, intramural and extramural growth bulging the serosa (note intact compressed subserosal tissue between arrows). IIA: small bowel GIST with a polypoid luminal (upper) and predominant extramural (lower) components. The lower component was separated from the abdominal cavity merely by a thin serosal covering. MB: this broad-based small bowel GIST showed exclusively extramural (pseudo-diverticular) growth without detectable luminal component. IIC: this huge gastric GIST presented as omental mass with minor adhesions to the external gastric wall.

Statistical analysis

Relationships between dichotomized histological types and clinicopathologic factors were examined for statistical significance using the log-rank test (significant association with p<0.05). Recurrence-free survival (= primary outcome measure) was determined by the appearance of tumor manifestation at peritoneal sites distinct from the primary tumor mass, either during surgery (synchronous peritoneal spread) or on follow up (metachronous peritoneal spread). Patients lost to follow-up were treated as censored cases based on the date they were last known to be alive. Recurrence-free survival curves were generated using the Kaplan-Meier method, and log-rank tests compared the distributions between the two groups.

Results

67 tumors from 67 patients with well documented gross features were recovered from a collective of more than 200 tumors at our departments (Table 2). These included 35 men and 32 women aged 25-86 yrs (mean, 64 yrs). None of the patients included had received a neo-adjuvant imatinib treatment and none of those with initially limited disease had an adjuvant treatment prior to tumor progression. Follow-up was available for 30 patients (mean, 29 months; range, 6-86 months). Further 10 patients with evident synchronous metastatic disease to the peritoneum at the time of primary diagnosis were included in the endpoint group (total 40 patients). Tumor rupture was observed in two cases (3%). Examples of tumor rupture and tumor-serosal relationship of the two macroscopic types are illustrated in Figure 2.

Table 2.

Clinicopathological and macroscopic pattern in gastrointestinal stromal tumors (n=67)

No	Age/Sex	GIST site	Size cm	Type	Risk Fletcher	Risk Miettinen	Follow-up in months (M)	Macroscopic type
1	63 F	Stomach body	1.5	Mixed	IR	Uncertain	44 MANED	IA
2	59 F	Proximal duodenum	1.2	Spindle	VLR	None	41MANED	IA
3	73/M	Stomach body	4	Mixed	LR	VLR	NA	IA
4	69 M	Stomach lower body	2.5	Mixed	LR	VLR	NA	IB
5	80 M	Antrum	2	Epithelioid	VLR	None	NA	IB
6	57 M	Fundus	2.4	Spindle	IR	IR	NA	IB
7	25 F	Antrum	3.5	Mixed	IR	IR	NA	IB
8	86 F	Stomach	4.5	Mixed	LR	VLR	26 MANED	IB
9	70 F	Stomach	4.5	Spindle	LR	VLR	32 MANED	IB
10	78 M	Stomach body	0.8	Epithelioid	VLR	None	NA	IB
11	52 F	Duodenum	3.7	Spindle	LR	LR	42 MANED	IB
12	73 F	Antrum	4.5	Spindle	LR	VLR	NA	IB
13	64 M	Stomach body	1.5	Spindle	VLR	None	57 MANED	IB
14	83 F	Antrum	4.8	Epithelioid	LR	VLR	28 M liver MTS 35 M DOD	IB
15	50 M	Stomach body	5	Epithelioid	LR	VLR	NA	IB
16	71 F	Fundus	5.8	Spindle	IR	LR	NA	IB
17	78 F	Appendix	0.5	Spindle	VLR	None	NA	IC
18	63 M	Stomach cardia	6.5	Mixed	HR	HR	6 M peritoneal MTS, 29 M DOD	IC
19	67 F	Small intestine	3.4	Spindle	LR	LR	20 MANED	IC
20	75 F	Stomach body	6.5	Spindle	IR	LR	14 MANED	IC
21	51 F	Stomach	6	Spindle	IR	LR	NA	IC
22	57 M	Fundus	4.5	Spindle	IR	IR	Recent case	IC
23	73 M	Jejunum	4	Spindle	HR	HR	Recent case, synchronous small serosal satellite	IIA
24	69 F	Jejunum	5.5	Spindle	IR	IR	Recent case	IIA
25	74/F	Fundus	28	Spindle	HR	HR	8 M ANED	IIA
26	74/M	Jejunum	>3	Spindle	HR	HR	Initial peritoneal MTS, 15 M DOD peritoneal+hep MTS under Glivec	IIA
27	84 M	Jejunum	7.5	Spindle	HR	HR	Simultaneous peritoneal sarcomatosis	IIA
28	73 F	Small intestine pelvic	7.2	spindle	IR	IR	NA	IIA
29	46 M	Stomach body	>5	Mixed Imatinib	HR	HR	Simultaneous peritoneal+liver dissemination	IIA
30	71 M	Small intestine	6	Spindle	IR	IR	15 M NED	IIA
31	79 M	Stomach body	7.5	Spindle	IR	LR	20ANID	IIA
32	65 M	Stomach body	6.5	Mixed	IR	IR	19 M ANED	IIA
33	55 M	Sigmoid	5.5	Mixed	HR	HR	Liver MTS 39 M 78 M AWD	IIA
34	85 M	Jejunum	12	Spindle	HR	HR	NA	IIB
36	75F	Ileum tumor “in Meckel"	8	Spindle	IR	IR	28 M, peritoneal + subcutaneous recurrence	IIB
37	59M	Small intestine	9	Spindle	IR	IR	19 M ANED	IIB
38	70F	Small intestine	1.5	Spindle	VLR	None	42 M ANED	IIB
39	65F	Small intestine	>3	Spindle	MTS	malignant	Died initially of disease Extensive MTS perit.	IIB
40	70M	Small intestine omentum	3.5	Mixed	HR	HR	Died initially of peritoneal sarcomatosis	IIB
41	69F	Jejunum + duodenum	2 tumors 5.8 each	Mixed	IR	HR	44 M ANED	IIB
42	54F	Rectum/ Douglas	4.5	Spindle	HR	HR	40 M ANED	IIB
43	64M	Stomach body	23	Mixed	HR	HR	7 M DOOC, no MTS	IIB
44	78M	Antrum	3.3	Spindle	LR	VLR	NA	IIB
45	52M	Rectum retrovesical	>4	Spindle	HR	HR	14 M pelvic recurrence	IIB
46	44F	Duodenum/ retroperitoneal	6	Mixed	IR	HR	16 M ANED	IIB
47	58F	Ileum	23	Mixed	HR	HR	Primary tumor rupture, Glivec, 24 M Recurrence	IIB
48	76M	Terminal ileum	5.3	Spindle	HR	HR	11 M ANED	IIB
49	62M	Small intestine	16	Spindle	HR	HR	24 M extensive peritoneal recurrence	IIB
50	84M	Stomach body	2 tumors 1.2 and 2	Spindle	VLR	None	NA	IIB
51	72M	Appendix tip	2.5	Spindle	LR	LR	NA	IIB
52	66F	Stomach	9.5	Spindle	IR	LR	NA	IIB
53	69M	Fundus	20	Spindle	HR	HR	NA	IIB
54	62F	Antrum	15	Epithelioid	HR	IR	NA	IIB
55	52F	Ileum	2.1	Mixed	LR	LR	NA	IIB
56	80M	Jejunum	4	Mixed	LR	LR	NA	IIB
57	52M	Stomach	2.8	Mixed	LR	VLR	NA	IIB
58	64/M	Ileum “tumor in Mickel”	5	Epithelioid	LR	LR	NA	IIB
59	67/F	Jejunum	13	Spindle	HR	HR	18 M: Multiple peritoneal MTS	IIB
60	73/F	Stomach	6	Spindle	IR	HR	19 M ANED	IIB
61	64F	Ileum	6.5	Spindle	HR	HR	Simultaneous sarcomatosis peritonei	IIB
62	65/M	Small intestine	6.4	Mixed	HR	HR	23 M ANED	IIC
63	76/M	Ileum	10	Spindle	IR	IR	Initial infiltr abd. wall+ ant. bladder roof	IIC
64	66M	Antrum	3	Epithelioid	LR	VLR	NA	IIC
65	72F	Small intestine	11	Spindle	HR	HR	Minute peritoneal MTS simultaneously, ruptured	IIC
66	71F	Stomach	12	Spindle	HR	HR	86 M ANED	IIC
67	72F	Small intestine	14	mixed	HR	HR	72 peritoneal recurrence Omentum MTS	IIC

VLR, very low risk; LR, low risk; IR, intermediate risk; HR, high risk; M, month; ANED, alive with no evidence of disease; MTS, metastasis; AWD, alive with disease; DOD, died of disease; NA, not available; DOOC, died of other cause.

Figure 2.

A. this type MB small bowel GIST presented with spontaneous old tumor rupture (seen as yellowish plaque-like depression) on the top of the mass (histology showed infarct-like necrosis). B. intra-operative view of simultaneous peritoneal spread from the same patient. C. histological section from another type II tumor showed microscopic penetration of the covering serosa (arrows). D. histological section from a type I tumor showed intact compressed subserosal tissue (upper).

Distribution of tumor types in relation to anatomic site, risk group and histological subtype

Tumors originated in the stomach (n=32; 48%), duodenum (n=3; 4%), jejunum/ileum (n=27; 40%) and large bowel/appendix (n=5; 7%). Based on the above defined gross classification, tumors were classified as type I (n=22) and type II (n=45). Of all, gastric GISTs are significantly associated with type I than type II (18 type I gastric vs. 14 type II gastric cases, p<0.001). On the contrary, intestinal tumors were predominantly of type II (31 type II intestinal vs. 4 type I intestinal cases, p<0.001). Based on the National Institutes of Health (NIH) consensus criteria (Fletcher et al, 2002) [5], tumors were classified as very low risk (n=6), low risk (n=17), intermediate risk (n=19) and high risk (n=25) for aggressive behavior. Applying this risk system, type I tumors were associated more frequently with a very low/low risk (n=14) than intermediate/high risk (n=8) contrasting with type II tumors (9 very low/low vs. 36 intermediate/high risk respectively). When the Miettinen's classification was used [1], 18 of type I tumors fitted the none/very low/low risk while only 4 tumors were of intermediate/high risk. The contrary was seen with type II tumors using the Miettinen's classification (34 intermediate/high versus 11 none/very low/low risk). The predilection of type II tumors for intermediate/high risk is of high statistic significance (p<0.001). Taken by histological subtype, type I tumors showed almost equal distribution into spindled (n=12) and mixed/epithelioid (n=10) types. On the contrary, a majority of type II tumors (n=29) were of the spindle cell type, but this was not significant (p≤0.435). The distribution of tumor types in relation to anatomic site, risk group and histological type is showed in Table 3.

Table 3.

Distribution of tumor types in relation to anatomic site, risk group and histologic subtype

Variable	Type I tumors	Type II tumors	p-value
gastric tumor	18	14	<0.001
non-gastric tumor	4	31	<0.001
VLR/LR (Fletcher et al)	14	9	<0.001
IR/HR (Fletcher et al)	8	36	<0.001
0/VLR/LR (Miettinen)	18	11	<0.001
IR/HR (Miettinen)	4	34	<0.001
Spindle	12	29	<0.435
Epithelioid/mixed	10	16	<0.435

0: no risk (benign in the Miettinen’ classification); VLR: very low risk, LR: low risk, IR: intermediate risk, HR: high risk

General frequency and site of metastasis in the study cohort

Of the 67 patients, 18 patients (27%) had evidence of peritoneal and/or hepatic spread either at the time of initial diagnosis or during follow-up. However, considering only those patients with available follow-up and those who initially have reached the primary endpoint (n=40), the frequency of malignant behavior would be 40%, thus being consistent with previous studies [1, 3, 19]. Ten of the 18 patients (55%) had metastatic spread at the time of initial diagnosis, mostly to peritoneal sites. Metas-tases were either confined to peritoneal sites (35%), limited to liver (5%) or they affected both sites (5%). Thus taken all together, peritoneal and liver metastasis occurred in 40% and 10% of the 40 cases, respectively.

Correlation of tumor macroscopic pattern with disease-free survival and pattern/site of progression

Of the 40 patients with available follow-up/metastatic disease, peritoneal progression was observed in 1/10 versus 15/30 patients with type I and type II tumors, respectively (p=0.032; Table 4). The rate of hepatic metastasis in patients with type I tumors (1/10; 10%) was similar to those with type II tumors (3/30; 10%; p=1). None of the patients with type I tumors had evidence of synchronous peritoneal metastasis at the time of surgery. Synchronous peritoneal metastases were observed only in type II tumors (n=10), 2 of them had hepatic progression at the same time. Metachronous peritoneal dissemination was seen in 6 patients, only 1 of them had type I tumor. Of the 16 patients with peritoneal progression, 14 had intestinal tumors and only 2 had gastric GISTs. The recurrence-free survival for the two macroscopic types is shown in the Kaplan-Meier analysis (Figure 3). In a total of 4 cases with liver metastases, 3 (75%) were patients with type II tumors, interestingly with type IIA tumors.

Table 4.

Correlation of macroscopic types with peritoneal metastasis in 40 GISTs

Gross pattern	Cases with follow-up	Synchronous peritoneal metastasis	Metachronous peritoneal metastasis	Peritoneal metastasis total
Type IA	2	0	0	0
Type IB	5	0	0	0
Type IC	3	0	1	1
Type IIA	9	4	0	4
Type IIB	16	4	4	8
Type IIC	5	2	1	3
Total	40	10	6	16

Figure 3.

Kaplan-Meier plot for the two types of GIST showed clear difference in the recurrence-free survival (marginal significance; p<0.057).

Discussion

GISTs differ from peripheral and retroperitoneal soft tissue sarcoma in that they do arise within hollow viscera of the GI tract. Accordingly, the deeper muscle layers and the peritoneal covering of the GI organs represent the sole physical barrier that prevents tumor dissemination to the abdominal cavity. Thus one should anticipate that involvement, penetration or disruption of these structures (barriers) represents a risk for dissemination in the same way common GI carcinomas do. To our knowledge, only a single previous study on a small series of GIST (n=25) has identified histological serosal penetration as an adverse prognostic factor but details of the methods used in that study were not available and the authors have used different risk classification system [21]. Thus, there exists to date no systematic analysis of the impact of the macroscopic growth pattern on the clinical course of GIST and, particularly, on the pattern of tumor recurrence. Lack of studies on this aspect in GIST pathology is probably explainable by the fact that most of the large previous studies have included predominantly clinically malignant referral cases, for which only limited or no detailed gross descriptions were available. In our experience and in the light of lacking standardized protocols for GIST specimen processing and reporting [25], pathologists generally do not make great effort to look for evidence of subtle serosal penetration at the time of gross description and during microscopic evaluation of GIST specimens, except for rare cases with overt tumor rupture or fragmentation. However, in the light of growing tendency to establish a standardized GIST reporting, the American College of Pathologists recently published guidelines and recommendations for GIST specimen evaluation and reporting [26]. In addition, a new TNM classification for GIST was included in the current version [27]. Unfortunately, both did not include any note on the significance of tumor rupture and/or serosal penetration in the T-classification and tumors have been classified solely on the basis of the same parameters (size, mitoses and site) used for conventional risks classification. While the role of overt tumor rupture has been pointed to by Joensuu in the revised NIH classification [7] and confirmed by recent series [20], detailed studies to identify additional risk factors relating to local tumor growth pattern and mechanisms of tumor spread in GIST without overt tumor rupture are still lacking. Thus, it is mandatory to perform larger prospective studies that take into consideration the macroscopic growth pattern of GIST and the tendency of large extramural GISTs to infiltrate adjacent organs [28].

In this study, we have analyzed the spectrum of all possible macroscopic tumor patterns/types in GISTs and tried to correlate them to patient outcome and pattern of disease progression in a series of well documented 67 GISTs recovered from retrospective GIST database at our institutions. Although limited by the lack of follow-up in several cases and its retrospective nature, this study demonstrated interesting points with regard to the natural course of GISTs that have not been addressed in previous clinicopathological series. Using the gross classification defined for the purpose of this study, a significant correlation was found between type I tumors and gastric location, lower risk category and paucity of peritoneal recurrence. On the other hand, type II tumors correlated significantly with intestinal location, higher risk category and frequency of peritoneal dissemination. Of interest, 62% of type II tumors with documented peritoneal spread showed evidence of peritoneal dissemination at the time of initial diagnosis/surgery. This observation underscores the necessity for scrutiny inspection of the peritoneal cavity during surgery for extramural and large GISTs, particularly of the small bowel. Detection of peritoneal seeding has a major impact on the patient's treatment and prognosis irrespective of the formal risk category of the tumor and is the mainstay of the indication for a life-long tyrosine kinase inhibitor therapy.

The rate of hepatic metastasis was similar in both type I (10%) and type II (10%) tumors. Interestingly, 3 of 4 tumors that gave rise to hepatic metastasis had a considerable luminal component suggesting that submucosal rather than extramural venous invasion might be responsible for liver metastasis in these cases but this remains currently speculative.

Our results may be of value with regard to prediction of metastatic sites in high risk GIST and hence would influence the strategies of follow-up in individual cases. The almost lacking peritoneal dissemination/recurrence in type I tumors can be explained by the fact that the presence of normal gut wall layer composed not only of the mesothelial layer but also of subserosal tissue and residual smooth muscle tissue represents an effective mechanical barrier against peritoneal contamination and this actually is the hallmark for distinguishing the two tumor types in this study. This is consistent with the well established role of serosal penetration as adverse marker of peritoneal recurrence in GI carcinomas. Admittedly, it was not possible to give a definitive statement with regard to microscopic serosal penetration in all of our cases. However, all of the tumors classified as type II tumors were covered solely by a thin serosal membrane at their peritoneal surface. Given the commonly larger size of these tumors, it is likely that such tumors would have minor serosal defects or microscopic tears related either to spontaneous mobility of tumors or during surgical manipulation. Admittedly, it would be difficult to thoroughly examine such tumors for focal microscopic serosal penetration given their commonly large capsular surface.

In summary, we described the gross spectrum of GIST, demonstrating that their diverse growth pattern not only of differential diagnostic relevance, but is also of potential value for predicting tumor recurrence and the site of metastatic spread. Further extended prospective studies on large series of site- and risk-matched cases are needed to clarify whether the gross type in GIST is an independent adverse parameter in their staging and prognosis.