Ronald P. DeMatteo is a consultant to and on the advisory board of Novartis. He is also a speaker for and has received honoraria from Novartis
Tumor mitotic rate, size, and location independently predict recurrence after resection of primary gastrointestinal stromal tumor (GIST)
Version of Record online: 12 DEC 2007
Copyright © 2007 American Cancer Society
Volume 112, Issue 3, pages 608–615, 1 February 2008
How to Cite
DeMatteo, R. P., Gold, J. S., Saran, L., Gönen, M., Liau, K. H., Maki, R. G., Singer, S., Besmer, P., Brennan, M. F. and Antonescu, C. R. (2008), Tumor mitotic rate, size, and location independently predict recurrence after resection of primary gastrointestinal stromal tumor (GIST). Cancer, 112: 608–615. doi: 10.1002/cncr.23199
- Issue online: 18 JAN 2008
- Version of Record online: 12 DEC 2007
- Manuscript Accepted: 21 AUG 2007
- Manuscript Revised: 6 AUG 2007
- Manuscript Received: 25 MAY 2007
- MRSG-04-027 from the American Cancer Society. Grant Number: P01 CA47179
- American College of Surgeons Oncology Group. Grant Numbers: CA94503, CA102613
- gastrointestinal stromal tumor;
Gastrointestinal stromal tumor (GIST) is the most frequent sarcoma of the intestinal tract and often shows constitutive activation of either the KIT or PDGFRA receptor tyrosine kinases because of gain-of-function mutation. Although the efficacy of tyrosine kinase inhibitors in metastatic GIST depends on tumor mutation status, there have been conflicting reports on the prognostic importance of KIT mutation in primary GIST.
A total of 127 patients were studied who presented to our institution from 1983 to 2002 with localized primary GIST and underwent complete gross surgical resection of disease. The majority of tumors originated in the stomach (58%) or small intestine (28%). By using polymerase chain reaction (PCR) and direct sequencing, a KIT mutation was found in 71% of patients and a PDGFRA mutation in 6%.
After a median follow-up of 4.7 years, recurrence-free survival was 83%, 75%, and 63% at 1, 2, and 5 years, respectively. On multivariate analysis recurrence was predicted by ≥5 mitoses/50 high-power fields, tumor size ≥10 cm, and tumor location (with patients having small bowel GIST doing the worst). In particular, a high mitotic rate conferred a hazard rate of 14.6 (95% confidence interval, 6.5–32.4). Specific KIT mutations had prognostic importance by univariate but not multivariate analysis. Patients with KIT exon 11 point mutations and insertions had a favorable prognosis. Those with KIT exon 9 mutations or KIT exon 11 deletions involving amino acid W557 and/or K558 had a higher rate of recurrence, whereas patients without a tyrosine kinase mutation had intermediate outcome.
In the absence of therapy with tyrosine kinase inhibitors, recurrence in completely resected primary GIST is independently predicted by mitotic rate, tumor size, and tumor location. Cancer 2008. © 2007 American Cancer Society.
Gastrointestinal stromal tumor (GIST) is the most common mesenchymal tumor of the intestinal tract. The tumor occurs typically in the stomach or small intestine, infrequently in the colon, rectum, or esophagus, and rarely outside of the gastrointestinal tract. A variety of mutations have been detected in GIST. The most common site involves KIT exon 11, whereas an exon 9 mutation occurs in approximately 5%.1 In about 3% to 5% of GISTs there is instead a mutation in the platelet derived growth factor receptor alpha gene (PDGFRA).2 In mice, targeted mutation of the KIT receptor has been shown to be sufficient for GIST development.3, 4
GIST has recently attracted widespread interest because of the development of effective targeted molecular agents against it. Over 80% of patients benefit from tyrosine kinase inhibitor therapy and the median survival from the diagnosis of metastatic GIST is now nearly 5 years.5 Whereas the outcome of targeted therapy in metastatic GIST has been shown to correlate with mutation status,6, 7 it is unclear whether mutation status predicts the likelihood of recurrence after resection of a primary GIST. Several investigators have reported that KIT mutation confers a poor prognosis in primary GIST.8–13 Because KIT mutations are known to be present even in small (<1–2 cm), incidental GISTs,14 we postulated that KIT mutation itself may not actually influence outcome after resection of a primary GIST. The aim of this study was to determine the relative impact of clinicopathologic factors on recurrence in a large series of surgically resected primary localized GIST treated at a single institution.
MATERIALS AND METHODS
From 1983 to 2002 there were 127 patients who presented to our institution with primary GIST without metastasis, underwent complete gross resection, and had adequate archival tissue for mutational analysis. Patient, tumor, and treatment data were collected prospectively and entered into our sarcoma database.
All tumors included in the study were re-reviewed by a single sarcoma pathologist (C.R.A.) and the diagnosis of GIST was confirmed by positive staining for KIT (CD117) protein, as previously described.15 Tumor morphology was classified as predominantly epithelioid or spindle-shaped. Mitotic rate was determined by counting the number of mitotic figures per 50 high power fields (HPF) and categorized as <5, 5–10, or ≥10.
Mutation analysis was performed as detailed previously.15 DNA was isolated from frozen tumor specimens using phenol chloroform extraction or from paraffin-embedded tissue using proteinase K digestion. One microgram of DNA was subjected to PCR using Platinum TaqDNAPolymerase High Fidelity (Life Technologies, Gaithersburg, Md). Oligonucleotide primers for KIT exons 9, 11, 13, and 17 or PDGFRA exons 12 and 18 have been published previously.15–17 Tumors were first tested for KIT exon 9 and 11 mutations. Tumors lacking a mutation were then tested for KIT exon 13 or 17 mutation and PDGFRA exon 12 and 18 mutations. The PCR products were identified by agarose gel electrophoresis using a 2% MetaPhor agarose gel (BioWhittaker Applications, Rockland, Me) and purified with the QIAquick PCR Purification Kit (Qiagen, Valencia, Calif) before sequencing. Every ABI sequence was compared with a National Center for Biotechnology Information (NCBI) Human KIT gene nucleotide sequence and blasted using an NCBI Standard Nucleotide Blast Search to determine the location and type of mutation within a particular exon. If no mutation was found initially, another portion of tumor was tested. Tissue collection and molecular analyses were approved by the Institutional Review Board.
Correlation between variables was assessed using the 2-sided Fisher exact test. For comparisons involving tumor location, 3 site categories and the 2-sided Pearson chi-square test were used. Actuarial recurrence-free survival from the date of surgical resection was calculated using the method of Kaplan and Meier. The relation of patient, tumor, and treatment characteristics to outcome was tested by univariate analysis using log-rank. SPSS statistical software (v. 11.5; Chicago, Ill) was used for univariate analysis. Multivariate analysis was performed with the Cox proportional hazards model. A P-value <.05 was considered statistically significant. SAS 9.1 (Cary, NC) was used for multivariate analysis.
The median age of the population was 67 years (range, 10–94) and there were 73 (57%) males. Tumor locations included the stomach in 74 (58%), small bowel in 35 (28%), rectum in 13 (10%), esophagus in 3 (2.4%), colon in 1 (0.8%), and an extraintestinal site in the pelvis in 1 (0.8%) (Table 1). There was 1 pediatric patient, a girl with multifocal gastric tumors. The median tumor size was 6 (0.3–50) cm. Most (74%) patients had a low mitotic rate. Cellular morphology was predominantly spindle-shaped in 112 (88%).
|Tumor location||Stomach||74 (58)|
|Small intestine||35 (28)|
|Tumor size||<5 cm||51 (40)|
|5–10 cm||37 (29)|
|≥10 cm||39 (31)|
|Mitotic rate||<5||94 (74)|
|KIT exon 11||85 (67)|
|Other DEL||17 (13)|
|KIT exon 9||4 (3)|
|KIT exon 17||1 (1)|
Partial gastrectomy was the most common (51%) operation. Five patients underwent pancreaticoduodenectomy, 5 a proximal gastrectomy, 4 an esophagectomy, 2 a total gastrectomy, 5 an abdominoperineal resection, and 3 a pelvic exenteration. Perioperative mortality occurred in 4 patients (3.1%), 1 of whom underwent operation primarily for an intraabdominal angiosarcoma. Positive microscopic margins occurred in 19 (15%) patients. No patient was treated with a tyrosine kinase inhibitor before developing recurrence.
KIT and PDGFRA Genotype
Of the 127 patients with localized primary GIST, 90 (71%) had a mutation in KIT (Table 1). Of patients with a KIT mutation, exon 11 was the most common site in 85 (94%). No patient had an exon 13 mutation. KIT exon 11 mutations were subtyped as deletions in 52 patients (61%), point mutations in 22 (26%), and insertions in 11 (13%). The KIT exon 11 deletions were further divided between those that resulted in an amino acid change or deletion at amino acid W557 and/or K558 (DEL557or8), and those that did not. In 29 (23%) patients a KIT or PDGFRA mutation could not be identified.
Recurrence Outcome and Univariate Analysis
With a median follow-up for survivors of 5.2 (0.02–12.8) years after resection of the primary tumor, 63 patients are alive without disease, 30 have died of disease, 22 died of other causes, and 12 are alive with disease. With a median follow-up for patients free of recurrence of 4.7 years, median recurrence-free survival was not reached with 83% recurrence-free at 1 year, 75% at 2 years, 63% at 5 years, and 60% at 10 years.
On univariate analysis, tumor location, size, and mitotic rate predicted recurrence-free survival. Patients with gastric GIST fared better than those with a small intestine or colon/rectum primary (Fig. 1). Tumor size was also an important predictor (Fig. 2). Patients with a mitotic rate of <5 did markedly better than those with ≥5 mitoses per 50 HPFs (Fig. 3). There was no difference in recurrence between patients with a mitotic rate of 5–10 versus those with ≥10 (P = .67). Age, sex, morphology, microscopic margins, and tumor rupture or intraabdominal hemorrhage did not predict recurrence. However, only 4 patients had tumor rupture. Positive microscopic margins occurred in 19 (15%) patients and were more likely in the rectum (38%) than the stomach (12%) or small intestine (9%).
The presence of any mutation or of any KIT mutation did not predict recurrence by univariate analysis (P = .93 and P = .75, respectively). However, the type of mutation was associated with recurrence-free survival (Fig. 4). We found that patients with KIT exon 11 DEL557or8 did worse than patients with other KIT exon 11 deletions (P = .04) or KIT exon 11 point mutations or insertions (P < .001). Patients with KIT exon 11 point mutations or insertions had a better prognosis than patients whose tumor had no mutation (P = .02). The 4 patients with KIT exon 9 mutations each developed recurrent disease within 27 months. Patients with a KIT exon 11 deletion other than DEL557or8, a PDGFRA mutation, or no mutation had similar recurrence-free survival (P = .82). The 1 patient with a KIT exon 17 mutation was alive without recurrence at 36 months.
When variables that were significant on univariate analysis were entered into multivariate analysis, factors independently associated with recurrence were mitotic rate ≥5, tumor size ≥10 cm, and location (Table 2). Patients with a small bowel (hazard rate 3.3) or colon/rectum (hazard rate 1.2) GIST did worse than those with stomach GIST. We did not enter KIT exon 9 mutations into the multivariate analysis as there were only 4 patients with this mutation. KIT exon 11 DEL557or8 mutation did not independently predict recurrence on multivariate analysis. This was explained by the finding that the 18 patients with stomach tumors having this mutation did worse than the other patients with stomach GIST (P = .002), but the 8 patients with colon/rectum tumors and this mutation did better than the other 6 patients with colon/rectum GIST (P = .02). For the 8 patients with small intestine tumors containing KIT exon 11 DEL557or8, there was a trend toward worse outcome (P = .08). When tumors in the colon/rectum were removed from the multivariate analysis there was a trend for tumors with KIT exon 11 DEL557or8 to have a worse outcome (P = .08, hazard ratio 2.11; 95% confidence intervals 0.9–4.8). Furthermore, when just patients with KIT exon 11 DEL557or8 were analyzed the outcome was similar regardless of tumor site (P = .90).
|Variable||No.||Univariate P||Multivariate P||Hazard ratio (95% CI)|
|Tumor size ≥10 cm||39||.004||.007||2.5 (1.3–4.8)|
|Small bowel||35||—||—||3.3 (1.5–7.2)|
|Mitotic rate ≥5/50 HPF||33||<.001||<.001||14.6 (6.5–32.4)|
|Absence of KIT exon 11 PM/INS||94||.002||NS||—|
|KIT exon 11 DEL557or8||35||.001||NS||—|
|KIT exon 9 mutations||4||<.001||NA|
We found that tumor mitotic rate, size, and location each independently predicted recurrence-free survival after complete removal of primary, localized GIST. The effect of mitotic rate was dramatic, with a hazard rate of 14.6 for those with ≥5 mitoses/50 HPF. Although several large series of completely resected GIST have identified tumor size10, 12, 18–20 and mitotic rate10, 12, 19–21 as prognostic variables, the influence of anatomic location has been unclear. It has been observed that gastric GIST generally has a more favorable course than small intestine GIST and thus location is commonly held to be a prognostic variable.22–26 However, tumor location has only been shown to be an independent predictor of outcome on multivariate analysis in 1 study.12 In particular, we found that patients with colon/rectum GIST had a high rate of recurrence, with only 20% free of recurrence after long follow-up. Although we grouped colon and rectum GISTs together, only 1 patient had a colon primary and therefore no definitive statements can be made about this site.
Several associations were identified between specific mutations and recurrence on univariate analysis. In particular, patients with KIT exon 11 point mutations or insertions had a favorable prognosis, whereas those with KIT exon 11 DEL557or8 and those with KIT exon 9 mutations had a poor prognosis. The rate of exon 9 mutation was low,3 consistent with a prior report.12 It has been previously reported that KIT exon 11 point mutations are associated with longer recurrence-free survival after surgical resection of GIST.27 We and others have previously reported the favorable prognosis of patients with KIT exon 11 insertions having internal tandem duplications at the 3′ end.15, 28 All but 1 of the 11 KIT exon 11 insertions in this series was an internal tandem duplication. We and others have also shown that patients with KIT exon 9 mutations are associated with worse prognosis.15, 29–31 The poor survival of patients with KIT exon 11 DEL557or8 was noted by Wardelmann et al.32 in 2003. This has subsequently been confirmed in 2 other series of resected primary GIST.12, 33 In the series by Martin et al, worse recurrence-free survival was seen for the subset of KIT exon 11 with deletions involving both amino acids W557 and K558 (DEL557and8).12 In the original series by Wardelmann et al.32 there was also possibly a stronger association of worse survival with DEL557and8 as opposed to deletions involving only 1 of the amino acids. In contrast, we did not see a difference in recurrence-free survival between the 26 patients with KIT exon 11 DEL557and8 compared with the 9 with deletions involving only amino acid 557 or 558. We found a unique interaction between tumor location and KIT exon 11 DEL557or8 with regard to recurrence that has previously not been reported. KIT exon 11 DEL557or8 mutations were associated with a worse prognosis in the stomach when compared with other stomach GISTs. A trend (P = .08) for worse outcome also existed for small bowel GIST with DEL557or8. However, within colon/rectum GIST, DEL557or8 was actually associated with an improved prognosis. Nevertheless, it seems that the presence of a KIT exon 11 DEL557or8 predicts an outcome that is independent of tumor site because recurrence-free survival was superimposable for the 18 stomach, 8 small bowel, and 8 colon/rectum GISTs with this mutation.
It has been suggested that GISTs with PDGFRA mutations are associated with more favorable prognosis.34 We did not find an association between PDGFRA mutations and recurrence in this study. It is possible that the strong association between PDGFRA mutations and gastric tumors accounts for this discrepancy, as gastric location was associated with a favorable prognosis. It has previously been noted that KIT exon 13 deletions are associated with malignant behavior30 and were also found to be associated with poor recurrence-free survival in 2 patients with KIT exon 13 deletions compared with 46 patients without this mutation in a series of primary GIST.27 We did not detect any KIT exon 13 mutations in the present report.
While there were correlations between specific mutations and recurrence-free survival on univariate analysis, we could not find an independent correlation with mutation on multivariate analysis. It is possible that associations between mutations and other prognostic variables and the relatively small number of patients with any given mutation prevented mutation from reaching significance. Similarly, for DEL557or8 the finding that colon/rectum tumors with this mutation actually had an improved prognosis favored site over mutation in our multivariate model, although tumors with this mutation appeared to have equivalent outcome regardless of their site.
Our finding that mutation analysis did not offer independent prognostic information for recurrence is in contrast to several previous reports.9, 10, 12, 19, 27, 33 Some of the discrepancy can be explained by small sample size19, 27 or the low prevalence of mutation9, 12, 19 in these studies. The low number of events in all these series, including the present report, complicates multivariate analysis. We had 42 recurrences, as did Martin et al, and no other report assessing the impact of mutation on recurrence has had as many. As certain specific mutations but not others correlated with recurrence on univariate analysis (Fig. 4 and Table 2), it is likely that the relative frequency of these mutations within any 1 series greatly influences the prognostic significance of groups such as ‘any mutation,’ ‘any KIT mutation,’ or ‘any KIT exon 11 mutation.’ Consequently, it is difficult to make comparisons across studies using what we now realize to be artificial designations. The interplay with site may explain why KIT exon 11 DEL557or8 was not an independent predictor on multivariate analysis in the present series. In the studies where KIT exon 11 deletions were found to be independent predictors of worse outcome, either site was not used in the multivariate analysis19 or there was a low number (3 [1.9%] and 2 [5.3%]) of colorectal tumors12, 33 compared with 14 (11%) in our series. If we excluded colon/rectum site as a variable in our analysis, KIT exon 11 DEL557or8 approached significance (P = .08) on multivariate analysis.
Overall, we were able to find a KIT or PDGFRA mutation in 77% of tumors. This is a relatively high mutation rate compared with other retrospective series.8–13, 19, 27, 33, 35 In contrast, the mutation rate of KIT or PDGFRA in recent, large, prospective trials of patients with advanced GIST has been approximately 86% to 87%.6, 7 Variability in mutation rate is likely due to methodologic differences. In the current study we detected mutations in 87% of 72 cases in which we had frozen tissue and only 64% of the 55 cases in which we had only paraffin tissue (P = .002). It is likely that degradation of DNA in archival paraffin tissue accounted for this discrepancy. We estimate that the use of paraffin tissue resulted in our inability to detect 12–13 mutations. Nevertheless, when we excluded the 35 cases where only paraffin tissue was available and no mutation was found, our findings were unchanged. We did not test for the recently identified PDGFRA exon 14 mutations, although we expect only 1 or possibly 2 cases should have harbored such a mutation.26
A better understanding of the prognostic factors in surgically resected primary GIST may allow for appropriate risk stratification that can be used for determining postoperative follow-up strategies and the need for adjuvant therapy. The possibility to delay or prevent recurrence with adjuvant treatment is even more important now that it is clear that acquired resistance to imatinib mesylate (Gleevec, Novartis, Basel, Switzerland) is a frequent event in metastatic GIST, which has a median time to progression of younger than 2 years.36–39 Although mutation did not independently predict recurrence-free survival, mutation is important in predicting response to tyrosine kinase inhibitors. Imatinib has proven to be highly effective in metastatic GIST40–43 with a 2-year survival of approximately 70% and a median survival of 58 months,5, 44 but the benefit varies based on mutation status. Notably, patients with a KIT exon 11 mutation have the best outcome. In 324 patients with advanced GIST being treated with 400 or 800 mg imatinib daily on the US phase 3 trial, patients with an exon 11 KIT mutation had an objective response rate of 67% and a median time to treatment failure of 576 days. Meanwhile, patients with an exon 9 mutation had a 40% response rate and those without a KIT or PDGFRA mutation had a 39% response rate and these groups failed at a median of 308 days and 251 days, respectively.6 In contrast, in patients with metastatic GIST who had mostly progressed on imatinib and were then treated with the multikinase inhibitor sunitinib malate (Sutent, Pfizer, New York, NY), those with KIT exon 9 mutations or no mutations fared best with 42% and 56%, respectively, having at least stable disease compared with 36% of those with KIT exon 11 mutations. Median survival for the 3 groups was 19.4, 20.9, and 5.1 months, respectively.
In summary, tumor mitotic rate, size, and location are independent predictors of recurrence-free survival for completely resected primary, localized GIST. Risk stratification after surgical resection should be based on these variables.
- 5Outcome of advanced gastrointestinal stromal tumor (GIST) patients treated with imatinib mesylate: Four-year follow-up of a phase II randomized trial Gastrointest Cancers Symp. 2006. Abstract 7., , , et al.
- 6Correlation of target kinase genotype with clinical activity of imatinib mesylate (IM) in patients with metastatic GI stromal tumors (GISTs) expressing KIT (KIT+) J Clin Oncol. 2005; 23(16S): 7., , , et al.
- 36Mechanisms of resistance to imatinib mesylate (IM) in advanced gastrointestinal stromal tumor (GIST). Proc Am Soc Clin Oncol. 2003; 22: 815. Abstract 3275., , , et al.