SEARCH

SEARCH BY CITATION

Keywords:

  • gastrointestinal stromal tumor;
  • incidence;
  • prevalence;
  • prognosis

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND

Recent breakthroughs regarding gastrointestinal stromal tumors (GIST) and their pathogenesis have redefined diagnostic criteria and have led to the development of molecularly targeted drug therapy. New treatment options mandate more accurate information regarding the incidence, prevalence, clinical behavior, and prognostic factors of GIST.

METHODS

All patients (n = 1460) who potentially had GIST diagnosed from 1983 to 2000 in western Sweden (population, 1.3–1.6 million) were reviewed, and 288 patients with primary GIST were identified. The incidence and prevalence of GIST were determined, and predictive prognostic factors, including current risk-group stratifications, were analyzed statistically.

RESULTS

Ninety percent of GISTs were detected clinically due to symptoms (69%) or were incidental findings at surgery (21%); the remaining 10% of GISTs were found at autopsy. Forty-four percent of symptomatic, clinically detected GISTs were categorized as high risk (29%) or overtly malignant (15%), with tumor-related deaths occurring in 63% of patients and 83% of patients, respectively (estimated median survival, of 40 months and 16 months, respectively). Tumor-related deaths occurred in only 2 of 170 of patients (1.2%) with very-low-risk, low-risk, or intermediate-risk tumors. The annual incidence of GIST was 14.5 per million. The prevalence of all GIST risk groups was 129 per million (31 per million for the high-risk group and the overtly malignant group).

CONCLUSIONS

GIST has been under recognized: Its incidence, prevalence, and clinical aggressiveness also have been underestimated. Currently existing risk-group stratification systems based on tumor size and mitotic rate delineate GIST patients who have a poor prognosis. Prognostication in patients with GIST can be refined using a proposed risk score based solely on tumor size and proliferative index. Cancer 2005. © 2005 American Cancer Society.

Gastrointestinal stromal tumors (GIST) are the most common nonepithelial tumors of the gastrointestinal tract. These tumors have been associated with conceptual as well as practical problems for several decades. Earlier lack of rigorous diagnostic criteria for GIST has been due to incomplete understanding of its origin and differentiation and has led to a highly variable nomenclature over the years. Prediction of clinical behavior based on morphologic features has been particularly difficult.1 Surgery alone or in combination with traditional chemotherapy or radiation therapy largely has been ineffective in treating the majority of patients with malignant GIST.2

Recent breakthroughs regarding the pathobiology and treatment of GIST include observations that GIST closely simulates the interstitial cells of Cajal (ICC) morphologically and immunophenotypically; detection of the tyrosine kinase receptor, KIT, in both ICC and GIST3, 4; and gain-of-function mutations in the KIT gene that play an early and important role in the development of GIST.4–7 These findings led to the remarkable development of the new molecularly targeted drug therapy with imatinib mesylate (Glivec®), which targets and inhibits the activated KIT tyrosine kinase receptor.8–10 This drug offers the first effective medical treatment for patients with malignant GIST.11–14

Epidemiological data virtually are nonexistent regarding the true incidence and prevalence of GIST. Reasons for this include the previous lack of well defined pathologic criteria for GIST, varying nomenclature for GIST over the past few decades, and the finding that nearly 60% of all GIST have been diagnosed as benign tumors or tumors of uncertain malignant potential and, thus, are not reported to national cancer registries.

The current study objectives were to analyze the incidence and prevalence of GIST in a defined population followed for a long period and to reassess the utility of previously reported prognosticators in GIST. Another objective of this study was to evaluate the usefulness of recently reported risk-group stratification schemes1 and potentially to refine prognostication.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Retrieval of Patients

We retrieved records on all potential patients with GIST who had been diagnosed from 1983 to 2000 within Västra Götaland, a province in western Sweden with a population of 1.3–1.6 million. Histopathologic examination of all patients was carried out in four hospital-based pathology laboratories, with each hospital covering a defined geographic area in Västra Götaland. Computerized files of histopathologic diagnoses according to the topographic and morphologic (T and M) coding system of SNOMED were used to retrieve all patients from 3 of the 4 hospitals from 1983 to 2000 and at 1 of the 4 hospitals from 1993 to 2000. T codes included all sites in the gastrointestinal tract; intraabdominal, retroperitoneal, mesenteric, omental, and pelvic areas; and the liver (to include patients who presented with metastases). M codes included all types of benign and malignant mesenchymal lesions, tumors of the autonomic and peripheral nervous systems, and benign and malignant tumors not further classified.

In total, 1460 patients matched the aforementioned T and M codes. In all patients, pertinent clinical information and pathology reports were reviewed. Nearly 500 patients were excluded immediately, because they obviously did not represent GIST. In the remaining 960 patients, slides were reviewed histologically; 310 of those patients were excluded after histologic review. The remaining 650 patients were studied in detail with a minimum number of immunohistochemical stains (see below).

Inclusion Criteria

Three criteria were required for a tumor to be classified as GIST: 1) tumor site in or adjacent to the gastrointestinal tract, mesentery, omentum, or retroperitoneum; 2) spindled and/or epithelioid histologic appearance compatible with GIST; and 3) unequivocal immunoreactivity for CD117. Three hundred ninety-eight of 650 analyzed patients fulfilled these criteria for GIST. Among those 398 patients, 95 were referred to Sahlgren University Hospital (SUH) for diagnosis and/or treatment from hospitals outside the Västra Götaland region, and 15 patients had a primary GIST diagnosed before 1983; those patients also were excluded from the population-based study. In total, 288 patients with primary GIST diagnosed between 1983 and 2000 remained. The current results were based on analysis of these 288 patients.

Morphologic Analysis

The following data were tabulated for each patient: histologic type (spindled, epithelioid, or a mixture of spindled and epithelioid tumor cells), degree of tumor cell pleomorphism (categorized as minimal, moderate, or severe), and mitotic rate (the number of mitoses per 50 high-power fields [hpf], with 1 hpf = 0.16 mm2). The patterns of immunoreactivity for CD117 and CD34 were recorded as diffuse versus focal, and the intensity of immunostaining was recorded as weak, moderate, or strong. The average proliferative index was determined by estimating the percentage of Ki-67-positive nuclei in 10 randomly selected hpf; the maximum proliferative index was determined by identifying “hot spots” (increased numbers of positively staining nuclei) and then evaluating the percentage of positive cells in a single hpf.

The following antibodies were used in the immunohistochemical analysis: CD117 (c-kit protooncogene product, polyclonal; code no. A4502; 1:50 dilution; Dako Corporation, Carpinteria, CA), CD34 (QBEND/10; monoclonal code no. M7165; 1:25 dilution; Dako Corporation), and Ki-67 (MIB1; monoclonal; code no. 7240; 1:200 dilution; Dako Corporation). Heat-induced epitope retrieval (microwave) was performed for Ki-67 (using Tris/ethylenediamine tetraacetic acid, pH 9.0) and CD34 and CD117 (using citrate buffer, pH 6.0).

Clinical Information and Follow-Up

Information regarding symptoms and treatment, dates of diagnosis, local recurrences, metastases and survival were retrieved from clinical records, autopsy reports and official population registries in all patients. Tumors were classified according to a consensus risk-group stratification system1 based on maximum tumor size and mitotic rate (per 50 hpf). The scheme includes four tumor risk groups, including a very-low-risk group (< 2 cm and < 5 mitoses/50 hpf), a low-risk group (2–5 cm and < 5 mitoses/50 hpf), an intermediate-risk group (< 5 cm and 6–10 mitoses/50 hpf or 5–10 cm and < 5 mitoses/50 hpf), and high-risk group (> 5 cm and > 5 mitoses/50 hpf or > 10 cm regardless of mitotic activity). In the current study, a fifth risk group called overtly malignant was added; this group included all patients who had tumors with proven metastases at initial diagnosis.

Statistical Methods

In the analysis of all prognostic factors, a Poisson regression15 was used to estimate a hazard function of the form exp(β0 + β1. x1 + …), where the β values were coefficients, and xi, i = 1, 2, …, were variables. The function was continuous as a function of all continuous variables, including time since diagnosis. The analysis was performed by a stepwise procedure, which, at the end, only included variables that were of significant importance in the multivariate context. Like the Cox regression model, the quantity exp(β) gives the hazard ratio when comparing two individuals who differ 1 unit of the current variable. The Poisson regression analysis resulted in a risk score, which was a linear combination defined by the variables and the β coefficients.

A special analysis was performed to elucidate how well the risk score captured the large variation in the risk of dying. The probability of dying within 5 years after diagnosis was calculated depending the risk score. The calculation of the percentile points was based on the piece-wise, linear transformation.

The survival function was calculated for the different risk groups according to Fletcher et al.,1 with the addition of an “overtly malignant” group. The follow-up was divided in intervals of 0.5 years, and the death rate within each interval was estimated by calculating the quotient between the number of events and patient years within the interval. Thus, a stepwise constant hazard function was obtained; and, by applying the mathematic relation between survival and hazard functions, the survival function was determined. The expected proportion of survivors was calculated taking age, gender, and calendar time into account using the aforementioned mathematic relation applied to hazard functions derived from official statistics of Sweden.

The incidence of GIST was calculated from the number of clinically detected primary GIST divided by the accumulated number of person years (17,862,068 person years) during the period 1983–2000. The prevalence was calculated from the incidence figures and the estimated death hazard function, taking into account risk groups and calendar time.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Histologic Review versus Primary Diagnosis

The original histologic diagnoses of all 288 GIST were as follows: GIST, 28%; leiomyoma, 34%; leiomyoblastoma, 13%; leiomyosarcoma (including epithelioid subtype), 18%; and other types of mesenchymal tumors (schwannoma, neurofibroma, various sarcomas), 7%. Forty-one percent of the 288 primary GIST originally were diagnosed as benign, and 42% as originally were diagnosed as potentially malignant or malignant. The biologic potential was not clearly stated in 17% of patients with GIST.

Detection of GIST and Clinical Features

One hundred ninety-nine GIST (69%) were detected due to symptoms. Sixty tumors (21%) were detected incidentally during surgery for other reasons. Thirty-two of those 60 tumors included those identified at surgery for other intraabdominal malignancies, such as colorectal carcinoma (15 tumors), gastric carcinoma (6 tumors), gynecologic carcinoma (3 tumors), pancreatic carcinoma (2 tumors), and urinary tract carcinoma (2 tumors). Twenty-nine GIST (10%) were detected incidentally at the time of autopsy (see Table 1).

Table 1. Detection, Gender, Age, Tumor Site, Tumor Size, and Risk-Group Distribution in 288 Patients with Gastrointestinal Stromal Tumors
DetectionNo. of patients (%)Gender (M/F)Median age in yrs (range)Tumor site (no. of patients)aTumor size (cm)Risk group: No. of patients (%)b
MedianMeanRange
  • M: male; F: female;

  • a

    Five patients with neurofibromatosis type 1 syndrome presented with multiple gastrointestinal stromal tumors: The data for the largest tumor were recorded.

  • b

    Risk groups were determined according to the groups reported by Fletcher et al., 2002,1 as follows: VL, very low risk; L, low risk; Int, intermediate risk; H, high risk; OM, overtly malignant.

Clinical-symptomatic199 (69)106/9367 (10–92)Gastric (120), duodenum (12), jejunum-ileum (50), colon (6), rectum (10), mesenterium (1)8.96.01.0–35.0VL: 5 (2.5); L: 62 (31); Int: 45 (22.5); H: 58 (29); OM: 29 (15)
Clinical-incidental60 (21)24/3674 (42–89)Gastric (23), duodenum (1), jejunum-ileum (32), colon (2), omentum (2)2.72.00.5–10.0VL: 27 (45); L: 23 (38.5); Int: 8 (13.5); H: 2 (3); OM: 0 (0)
Autopsy-incidental29 (10)14/1578 (48–90)Gastric (27), jejunum-ileum (2)3.41.50.5–10.0VL: 16 (55); L: 10 (34.5); Int: 2 (7); H: 1 (3.5); OM: 0 (0)
All288 (100)144/14469 (10–92)Gastric (170), duodenum (13), jejunum-ileum (84), colon-rectum (18), other (3)7.04.50.5–35.0VL: 48 (16.5); L: 96 (33.5); Int: 55 (19); H: 60 (21); OM: 29 (10)

Treatment

All but 1 of 259 patients with clinically detected GIST underwent surgery. Surgical removal of the tumor was complete in all 117 patients with very-low-risk GIST and low-risk GIST, in 52 of 53 patients with intermediate-risk GIST, in 48 of 60 patients with high-risk GIST, and in 5 of 29 patients with overtly malignant GIST. Patients with overtly malignant GIST presented with peritoneal, omental, and/or mesenteric metastases (21 patients); liver and peritoneal metastases (4 patients); or solely liver metastases (4 patients). Fourteen patients who presented with metastases or who later developed metastases were treated with highly variable chemotherapy protocols. Two patients received combined radiotherapy and chemotherapy.

Clinical Follow-Up

The correlation between risk group and outcome is detailed in Table 2 and is illustrated in Figure 1. The estimated median survival for the 259 patients with clinically detected GIST (including the 60 incidentally detected GIST during surgery for other tumors) was 129 months, compared with a median survival of 214 months for the age-matched and gender-matched, normal population.

Table 2. Risk Groups and Clinical Outcomes in 259 Patients with Clinically Detected Gastrointestinal Stromal Tumors
Risk groupNo. of patientsPersistent tumor diseaseaRecurrent tumor diseasebNo. of tumor-related deaths (%)cEstimated medial overall survival (yrs)
  • a

    Patients who had residual tumor after surgery.

  • b

    Patients who had local recurrence and/or metastases after complete surgical removal.

  • c

    Deaths due to persistent, recurrent, or metastatic gastrointestinal stromal tumor (GIST).

  • d

    This was a 73-year-old man with a 4-cm gastric GIST who died 5 years later with liver metastases.

  • e This was a 77-year-old woman with a 3.8-cm gastric GIST who died 28 months later with local recurrence and liver metastases.

Very low320/320/32 0/32 (0)> 16
Low850/852/85 1/85d (1)> 16
Intermediate531/531/52 1/53c (2)14.2
High6012/6030/4838/60 (63)3.4
Overtly malignant2924/295/524/29 (83)1.4
Total25937/25938/22264/25910.9
thumbnail image

Figure 1. The estimated overall survival in 259 patients with clinically detected gastrointestinal stromal tumors is compared according to risk group with an age-matched and gender-matched, normal population.

Download figure to PowerPoint

Thirty-eight patients had recurrent disease after undergoing complete surgical resection of their tumors; 5 patients had local recurrences, and 33 patients had metastases, including combined liver and peritoneal metastases in 6 patients, solely liver metastases in 11 patients, and peritoneal, omental, and mesenteric metastases in 6 patients). Two of those patients also developed bone metastases, and one patient developed lung metastases.

Statistical Analysis

The following factors were statistically significant with regard to decreased overall survival: tumor size (P < 0.001), degree of cellular pleomorphism (none, minimal, or moderate vs. severe; P < 0.001), mitotic rate (< 2 mitoses/50 hpf, 2–5 mitoses/50 hpf, 6–10 mitoses/50 hpf, or > 10 mitoses/50 hpf; P < 0.001), and average and maximum Ki-67 proliferative index (percentage as a continuous variable; P < 0.001). Compared with the normal population, there was a statistically significant decrease in overall survival for all risk groups combined (P < 0.001), for the low-risk group (P = 0.0039), and for the high-risk and overtly malignant risk groups (P < 0.001). There was no significant decrease in survival for the very-low-risk group or the intermediate-risk group compared with the normal population.

It was found that neither gender nor tumor site (gastric vs. small intestine and colon) correlated with a decrease in overall survival (P > 0.05). However, the estimated risk of dying was 10% less for patients who had gastric GIST compared with the risk among patients who had small intestinal and colon GIST. There was no significant correlation between survival and histologic tumor type (spindled, epithelioid, or mixed types), CD117 immunoreactivity (weak vs. moderate vs. strong and focal vs. diffuse immunostaining), and CD34 immunoreactivity (positive vs. negative).

In the stepwise multiple analysis, only 2 factors were identified that independently influenced survival: maximum Ki-67 proliferative index (P < 0.00001) and tumor size (P = 0.0053). Statistical calculations were performed to illustrate the influence of maximum proliferative index on survival given a fixed tumor size and the influence of tumor size on survival given a fixed maximum proliferative index (Figs. 2, 3). Based on an analysis of tumor size, maximum proliferative index (%Ki-67 max), and survival, a risk score or predictor was determined: 0.0486 (tumor size in cm) + 0.0491 (%Ki-67max).

thumbnail image

Figure 2. The estimated proportion of survivors among 251 patients with clinically detected gastrointestinal stromal tumors is illustrated according to the Ki-67 maximum percentage (Ki 67 max%). In this estimate, tumor size was fixed at 5 cm. Each percentage point increase in Ki 67 max% was associated with a 5% increased risk of dying.

Download figure to PowerPoint

thumbnail image

Figure 3. The estimated proportion of survivors among 251 patients with clinically detected gastrointestinal stromal tumors is illustrated according to tumor size. In the estimate, the Ki-67 maximum percentage was fixed at 1%. Each 1-cm increase in tumor size was associated with a 5% increased risk of dying.

Download figure to PowerPoint

Because both coefficients are nearly identical (0.0486 and 0.0491), a simple risk score can be calculated. The probability of dying within 5 years, depending on the percentile point of the risk score (tumor size in cm + %Ki-67max) and age at diagnosis, is illustrated in Figure 4.

thumbnail image

Figure 4. The estimated probability of dying within 5 years in 251 patients with gastrointestinal stromal tumors is illustrated according to various risk scores (tumor size in cm and Ki-67 maximum percentage) and percentiles.

Download figure to PowerPoint

Up to the median percentile (50%), corresponding to a risk score of 7, the estimated survival does not differ significantly from the normal age-matched and gender-matched population. Risk scores > 7 are associated with an increased risk of dying within 5 years, with the risk rapidly increasing with increasing scores. The overall survival of patients with risk scores ≤ 7, statistically, differs significantly from the survival of patients with risk scores > 7 (P < 0.0001). The observed clinical courses for 251 patients given a risk score (tumor size [in cm] + %Ki-67max) of ≤ 7 and > 7 are illustrated in Table 3.

Table 3. Correlation between Gastrointestinal Stromal Tumor Risk Scores and Clinical Outcomes in 251 Patients with Clinically Detected Gastrointestinal Stromal Tumorsa
GIST risk score (size + Ki67 max)bNo. of patientsNo. of patients (%)Estimated medial overall survival (yrs)
Persistent tumor diseaseRecurrent tumor diseaseTumor-related deaths
  • GIST: gastrointestinal stromal tumors Ki67max: maximum Ki67 proliferative index.

  • a

    In 8 of 259 clinically detected GISTs, Ki67 immunostains could not be performed.

  • b

    The GIST risk score is the maximum percentage of Ki67-positive tumor cells plus the maximum tumor size (cm).

  • c

    This is the same patient in the low-risk group indicated under “No. of tumor-related deaths” in Table 2.

≤ 71232/123 (1.6)3/121 (2.5)1/123 (0.8)c14.3
> 712833/128 (26.0)35/95 (37.0)60/128 (47.0)6.0

Incidence and Prevalence

The annual incidence of clinically detected primary GIST within the Västra Götaland region was estimated at 14.5 per million inhabitants (95% CI, 12.8–16.4 per million). The annual incidence of GIST did not differ significantly over time. The prevalence of GIST was estimated to be 129 per million. For the various risk groups, prevalence was estimated as follows: very low risk, 22.2 per million; low risk, 51.9 per million; intermediate risk, 24.2 per million; high risk, 22.2 per million, and overtly malignant, 8.7 per million.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

GIST traditionally has been viewed as a rare, enigmatic, and unpredictable tumor capable of aggressive behavior, and surgery has been the mainstay of treatment. The previous lack of understanding of the origin of GIST, problems in distinguishing GIST from other mesenchymal tumors, difficulties in predicting the behavior of GIST, and limited treatment options have perplexed physicians for decades. Recent discoveries regarding the close relation of GIST to the pacemaker cells of the gut, its pathobiology, and critical molecular genetic events in its development have reawakened an interest in GIST. These findings have led to the development of a highly specific and effective, targeted therapy with imatinib mesylate, representing a paradigm shift in therapy for malignant disease.16 Thus, accurate histologic diagnosis, correct nomenclature, and accurate prediction of prognosis are essential for the current treatment of patients with GIST. The true incidence, prevalence, and biologic spectrum of GIST based on long-term follow-up studies also have become increasingly important.

A prerequisite for this first population-based study was a clear definition of GIST. The recent redefinition of histologic criteria for the diagnosis of GIST—a spindled and/or epithelioid mesenchymal tumor of the gastrointestinal tract with unequivocal KIT immunoreactivity1, 3, 17—now is widely accepted. The diverse terminology reflecting various views on the differentiation of GIST over the past few decades required retrieving a wide range of diagnoses for review. The finding that two-thirds of the tumors in the current study had been classified as smooth-muscle tumors (leiomyoma, leiomyoblastoma, and leiomyosarcoma) and only 28% as GIST is not surprising, in that the patients cases in our study were diagnosed between 1983 and 2000. During that period, views regarding the histogenesis of GIST slowly shifted from one of smooth-muscle origin to that of a less differentiated stromal cell. Predictably, smooth-muscle tumors were diagnosed less frequently as the concept of GIST became increasingly accepted. From 1995 to 2000, two-thirds of the GIST in the current study originally were diagnosed as GIST.

The recent explosion of information regarding GIST over the past few years is attributable largely to the detection of KIT immunoreactivity in GIST and the critical role of KIT in its pathogenesis. Although awareness of GIST has been increased and its diagnosis facilitated, we suspect that its malignant potential continues to be underestimated. Only 42% of GIST had been reported to national and regional cancer tumor registries, as illustrated in the current study; 12% of patients with high-risk GIST had been misdiagnosed with benign tumors.

The calculated annual incidence and prevalence of GIST (14.5 per million population and 129 per million population, respectively) must be viewed as minimum values. Although we included a wide spectrum of diagnoses in the retrieval of our patients from computerized data bases, we cannot exclude the possibility that GIST in other sites and with other histologic diagnoses were overlooked. In addition, our estimates do not include rare KIT-negative tumors, which we believe are true GIST (< 2% of patients; findings in agreement with other studies).1 The prevalence values for GIST will increase quickly in the postimatinib era, particularly in the high-risk and overtly malignant groups (with a combined prevalence of 31 per million population in the current study).

There are no earlier population-based studies regarding GIST. The National Cancer Institute Surveillance, Epidemiology, and End Results data from 1995 report that sarcomas as a group account for only 2.2%, 13.9%, and 0.1% of gastric, small intestinal, and colorectal malignancies, respectively.18 Our findings of a surprisingly high incidence of GIST agree with a previous “population-based sample estimation” from southern Finland that reported an annual incidence of GIST of 10–20 per million population.17

The current study also showed that small and clinically insignificant GIST frequently are detected as incidental findings at autopsy and during abdominal surgery for other reasons. The true incidence of incidental, asymptomatic GIST could not be evaluated in this analysis. Twenty-nine GIST in our study were detected among some 14,000 autopsies, suggesting an incidence of 2 per 1000 population. We suspect that the true incidence is much higher. The predominance of gastric GIST in the autopsied patients (95%) may reflect closer scrutiny of the stomach than the bowel at the time of autopsy. In addition, there probably is a tendency to examine larger tumors found at autopsy histologically. The predominance of small intestine GIST and colon GIST among incidentally detected tumors at surgery may reflect closer scrutiny of these organs during abdominal surgery.

Numerous studies have attempted to delineate criteria to distinguish benign GIST from malignant GIST.1 It has been found that tumor size and mitotic rate most consistently are the strongest predictors.1, 17, 19, 20 However, there are occasional reports of very small and mitotically inactive GIST with a malignant course.1, 19

We found that the recent suggestion to divide GIST into very-low-risk, low-risk, intermediate-risk, and high-risk groups,1 with our addition of an overtly malignant risk group, accurately predicted outcome with regard to both recurrent tumor and survival. Thus, among 170 patients with very-low-risk, low-risk, and intermediate-risk GIST, only 1 patient had an unresectable tumor, and 3 patients developed recurrent tumors (2 metastases and 1 local recurrence). In contrast, among 89 patients with high-risk or overtly malignant GIST, 36 patients had unresectable tumors, and 35 patients developed recurrent tumors. The extremely low recurrence rate among the patients with very-low-risk, low-risk, and intermediate-risk GIST may be related to the fact that 58 of 170 patients had tumors that were detected incidentally at surgery. Sixty-three percent of patients in the high-risk group and 83% of patients with overtly malignant tumors died due to GIST: The estimated median survival was 40 months and 16 months, respectively. In contrast, only 1% of patients in the very-low-risk, low-risk, and intermediate-risk GIST groups died of tumor. The significantly decreased overall survival for patients in the low-risk GIST group (P = 0.0039) can be explained by the fact that 24 of 85 tumors in this group were detected incidentally during surgery for other intraabdominal malignancies.

The current risk-group stratification according to Fletcher et al.1 does not take into account the possible influence of tumor site. It has been suggested that tumors of the small intestine have a more aggressive course than tumors of the stomach.17, 19 In the current analysis, tumor site did not correlate significantly with overall survival, although the patients who had gastric tumors carried a 10% lesser risk of dying than patients who had tumors of the small intestine. Another recently suggested GIST risk-group stratification system takes tumor site into account as well as tumor size and mitotic rate, dividing GIST into probably benign, uncertain or low malignant potential, and probably malignant.19 We also evaluated the predictive value of this risk-group stratification system in our study. Although the probably malignant group had a statistically significantly decreased overall survival (P < 0.001) compared with the normal population, the other 2 risk groups did not.

The degree of pleomorphism was correlated with survival, as observed previously.20 Only 1% of the very low, low, and intermediate risk-group tumors showed severe pleomorphism, in contrast to nearly 50% of those in the high-risk tumors and overtly malignant tumors.

Only two independent prognostic predictors (equally important) were identified in the statistical analysis: tumor size in cm and maximum proliferative index (percentage). The findings suggest that the Ki-67 proliferative index, particularly the identification of “hot spots,” has a stronger predictive value than mitotic rate. The Ki-67 index was reported previously as a strong prognostic indicator in GIST—in some studies: superior to the mitotic rate,21, 22 but inferior to the mitotic rate in other studies.23, 24

The simple risk score for GIST based on the sum of the tumor size and the Ki-67 maximum proliferative index effectively stratifies patients with GIST into one group with an excellent prognosis and one group with a poor prognosis. Among the patients with risk scores > 7, 63% had persistent or recurrent GIST, and 47% died of tumor. This is in contrast to patients who had predictive scores ≤ 7, only 4% of whom had persistent or recurrent tumors, and < 1% died of tumor. From a practical standpoint, any GIST > 6 cm in greatest dimension, regardless of the patient's proliferative index, and any GIST with a maximum Ki-67 proliferative index ≥ 5%, regardless of tumor size, should be regarded as potentially malignant.

No previous studies clearly identified the relative proportions of the various risk groups in the population of patients with GIST. For obvious reasons, the proportion of patients with clinically aggressive GIST has been overestimated in studies of patients with GIST who underwent surgical treatment at referral centers.2, 25 In unselected series, it has been estimated that only one-fifth of all GIST were clinically malignant.17 The current population-based study shows that almost half (87 of 199 tumors; 44%) of all clinically detected, symptomatic GIST belong to the high-risk group or the overtly malignant group and that two-thirds (125 of 198 tumors) have a risk score ≥ 7, which is associated strongly with malignant behavior. We believe that the proposed prognostic risk score based solely on tumor size and proliferative index is simpler and potentially more useful than existing stratification schemes for patients with GIST, particularly with regard to designing future treatment protocols.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES