Rapid progress has been made in the diagnosis and targeted treatment of gastrointestinal stromal tumors (GIST) since the identification of c-kit gene aberrations in GIST in 1998.1 Gain-of-function mutations of either c-KIT (85–90%) or, as shown more recently, PDGFR (4%) are supposed to play a critical role in tumorigenesis.2, 3 Moreover, they serve as a sensitive diagnostic marker for the differentiation between GIST and other soft tissue tumors and can be targeted with the tyrosine kinase inhibitor imatinib.4, 5
Until recently, surgery has been the only treatment approach for GIST, resulting in 5-year survival rates of 48–54% for resectable disease. In patients with metastasized, irresectable or locally recurrent disease, the outcome has been poor with a median survival of 12–19 months and a 5-year survival rate of <5–10%.6 “Classical” chemotherapeutic treatments yielded objective responses in only a few cases (overall response rate [OR] =0–5%) and are not supposed to have a positive prognostic impact for these patients.7
Imatinib, a selective receptor tyrosine kinase inhibitor, has been shown to effectively inhibit activated KIT and PDGFR in vitro.8 In recent Phase II/III trials, treatment with imatinib in patients with irresectable GIST yielded objective responses in about 50% and disease stabilization (stable disease) in 30–40% of the patients, respectively.9, 10 The corresponding 2-year survival rates are in the range of 70–80%, suggesting a marked improvement in treatment outcome as compared to anecdotal data from classical chemotherapy trials of the pre-imatinib era.11, 12, 13 Interestingly, after treatment with imatinib for advanced or metastatic disease, individual patients are now being observed who are eligible for secondary surgical resection. So far, no data have been available to show a potential prognostic benefit from secondary resection, and there are uncertainties with regard to the timing as well as the extent and outcome of resection.
To address such questions as resectability rate, extent of resection, pathological response and postoperative follow-up after preoperative/neoadjuvant imatinib, we analyzed 90 patients with metastatic GIST of whom 12 patients were forwarded to secondary resection after response induction to preoperative imatinib.
Material and methods
Between December 2001 and June 2004, 113 patients with GIST were consulted at the Medical Oncology units of the West German Cancer Center in Essen (n = 54), the University of Tübingen (n = 31) and the University of Hamburg (n = 28) with 90 of these patients showing metastatic or irresectable locoregional disease. Patients with histologically confirmed, metastatic or non-resectable disease who had shown a favorable clinical response to imatinib were evaluated by a surgeon and underwent resective surgery if possible. Resection was carried out as soon as the maximum radiological response to the treatment with imatinib was observed. Imatinib treatment was continued if possible and follow-up was carried out every 2–3 months. Informed consent was obtained following the institutional guidelines.
To evaluate the feasibility of the resection of residual disease, all patients who underwent resection were analyzed for previous oncological treatments, the extent of disease before and after imatinib treatment, type and result of surgery, post-operative complications, histopathological changes, time to progression and survival (Tables I,II). In addition, all patients with metastatic GIST who did not undergo resection were analyzed retrospectively for disease localization and response to imatinib. Actuarial survival curves and comparisons for responding patients (resected vs. non-resected) were calculated from the starting date of imatinib therapy according to the Kaplan-Meier product-limit method and the log-rank test, using statistical software (SSPS software; version 11.0; SPSS Inc., Chicago, IL).
Table I. Clinical Data and Tumor Characteristics in 90 Patients with Metastasized or Unresectable Locally Advanced GIST
Median follow-up since initiation of Imatinib (months)
Median follow-up since first diagnosis of GIST (months)
Six females and 6 males, representing 13% of the series of 90 patients with metastatic disease, were forwarded to secondary resection. The median age at imatinib start was 60 years (range = 33–75 years). Primary tumor sites were the stomach (n = 5), small bowel (n = 5) and retroperitoneum (n = 1). In one patient, the primary site of the tumor could not be determined but was suspected to originate from the colon. Tumor-related major surgery (laparotomy) before imatinib therapy had been carried out in 10 of 12 patients with a median number of 2 laparotomies (range = 0–4). Five patients had already undergone resection of metastases before treatment with imatinib. More than one metastatic site was found in 3 patients. Localizations of metastases were the liver (n = 9) followed by the peritoneum (n = 5) and the retroperitoneum (n = 1). Median diameter of the largest metastasis before treatment with imatinib was 5.4 cm (range = 3–16.4 cm). Six of twelve patients had received at least 2 chemotherapeutic regimens before treatment with imatinib. In 7 patients with metachronous metastases the median time between complete resection of the primary GIST and occurrence of first metastases was 15 months (range = 6.9–104 months). Synchronous metastases, defined as occurrence within 3 months of initial diagnosis, were observed in 5 patients.
Seventy-eight patients with metastatic or locally advanced disease did not undergo resection (Table 1). Median age at initiation of imatinib in this group of patients was 62 years (range =25–81). Most common metastatic sites were liver (68%), peritoneum (46%) and extraabdominal sites (23%).
Immunohistochemistry was carried out on both pre- and posttreatment specimens if available. Sections were stained for c-KIT as described previously14 as well as CD34 and various proliferation/differentiation markers such as MIB-1, BCL-2, S-100 and actin. Mitoses were counted per 10 HPF (high power fields) or, if rarely detected, in counts per 20 HPF. The percentage of viable cells in the postoperative specimen was estimated by screening several representative sections of the resected specimen(Table III).
Table III. Pathological Findings in 12 Patients with Metastasized GIST Before A Treatment and After A Clinical Response with Imatinib
% Vital tumor
GIST, with nodular hypercellular areas, in part fatty and myxoid degeneration, scattered inflammatory cells, focal hemorrhage
Near total stromal degeneration, few scattered residual tumor cells
Spindle cell type
Focal areas of hypercellular, vital tumor cells, areas ofcomplete necrosis
Imatinib treatment in unresected patients
In the group of 78 patients who did not undergo resection 38 patients showed (51%) complete or partial remissions to the treatment with imatinib, “no change” was observed in 21% and progression to therapy in 23% of the patients. As of the last follow-up, progression during imatinib treatment was observed in 48% of all 78 patients and in 29% of the patients with initial response (Fig. 1).
In the group of resected patients preoperative treatment consisted of a daily imatinib dose of 400 mg (−600 mg) for a median duration of 12.2 months (range = 6.1–25 months). Partial remissions were observed in 11 of 12 patients (according to WHO criteria, 50% reduction of the sum of the products of the biperpendicular diameters of measurable lesions); one patient showed a minor remission. 18FDG-PET scans were carried out in 8 of 12 patients showing increased 18FDG utilization of metastases in 6 of 8 patients before the start of the treatment. Areas of increased tracer uptake diminished completely after the start of imatinib in all 6 patients. The median number of metastases was 7.5 (range = 2–12) before and 2 (range = 0–12) after treatment with imatinib as measured by CT. For evaluation of resectability, pre- and post-therapeutic radiographic and functional imaging studies were reviewed by a surgeon and decisions depended upon the discretion of the responsible physicians. In general, liver tumors were assessed to be resectable if all tumors, regardless of number and size could be removed completely. The minimum amount of hepatic parenchyma necessary to preserve sufficient hepatic function was assumed to be at least 0.5% of total body weight, provided that the remaining parenchyma had no impairment of arterial and portal venous blood supply and a fully preserved hepatic venous outflow and biliary drainage.
Overall, 14 surgical procedures were carried out on 12 patients (Essen, 9; Hamburg, 2; Tübingen, 1). These procedures consisted of 7 hepatic metastasectomies (segmentectomies in 6 patients, 1 extended hemihepatectomy), resection of peritoneal lesions in 5 patients, resection of retroperitoneal lesions in 1 patient and laser-induced thermo therapy (LITT) in another patient. Complete resection of macroscopic disease (R0) was achieved in 11 patients with no evidence of residual disease in radiographic/functional imaging studies after surgery (Fig. 5).
In one patient, intraoperative ultrasound showed more lesions than had been diagnosed in preoperative imaging studies. To completely remove all manifestations a hemihepatectomy with subsegment resections of the contralateral lobe would have been necessary, which did not seem feasible for the patient. Therefore, only a diagnostic resection of a peripheral hepatic metastasis was carried out. Treatment with imatinib was resumed and the patient was reassessed 12 months later. Due to additional shrinkage of metastases, complete resection could be achieved by using only multiple subsegment resections. In a second patient, only incomplete resection of hepatic metastases was achieved and the 3 remaining metastases were treated by LITT. A control biopsy of the ablated areas showed no residual tumor cells.
In 2 patients, treatment with imatinib resulted in a complete radiological and macroscopical remission of the liver metastases and resections were confined to the residual extra-hepatic manifestations.
Significant postoperative complications occurred in 3 patients and consisted of prolonged wound healing due to local infection and fascial dehiscence (2/12), one bile leak requiring re-laparotomy, one abscess of the abdominal wall and one prolonged hospitalization due to postoperative pain, diarrhea and weight loss of 20%. Imatinib treatment was usually resumed as soon as the patients were started on solid food again. In this small series of patients there was no evidence for increased toxicity of imatinib after surgery or unusual wound complications.
Histopathological examination of tumor specimens showed that most areas of the metastatic lesions consisted of extensive stroma hyalinization, myxoid degeneration, and often scattered inflammatory cells. Focal hemorrhage as well as hemosiderin deposits were found in 3 patients. Signs of fatty necrosis were found in one patient and were otherwise not observed.
All but one specimen contained viable tumor cells within a median of 5% (range = 0–50%) of the areas examined (Table III). Proliferation markers such as mitotic count, MIB-1 or BCL-2 expression were evaluable in both pre- and posttreatment specimens in 9 of 12 patients. A significant reduction of these markers in the posttreatment specimens was found in 6 of 9 patients; in 3 patients a similar proliferative activity as in the pretreatment specimen was observed in the area of viable tumor cells. Patterns of residual cells consisted of diffusely spread viable single cells that were found within a dense hyalinized matrix as well as small foci of tumor cells; or large, nodular and hypercellular residues. KIT expression in the resected specimen was found in 9 of 11 patients who were positive before treatment with imatinib. Loss of CD34 immunoreactivity was seen in one patient. Three of nine patients did not show detectable levels of KIT of which 2 patients had obviously lost c-KIT expression and one patient did not show significant KIT expression before initiation of imatinib treatment as described earlier.14 These cells were spindle-shaped cells with positive staining for CD34, however, and in the context of this analysis, were considered to represent residual GIST cells (Figs. 2–4).
In 10 of 12 patients, imatinib therapy was continued after recovery from surgery with a median treatment duration of 16.9+ months (6–34+ months); 2 patients refused postoperative treatment. Ten of twelve patients have remained continuously free of progression after surgery with a median follow-up of presently 16.9 months (range = 1–34) since resective surgery and 29.8 months (range = 17–41) since initiation of imatinib therapy.
One patient with initially hepatic and widespread peritoneal disease was diagnosed with a localized, peritoneal relapse 6 and 27 months after the first resection while on treatment with imatinib. After 2 successful re-resections the patient recently showed no evidence of disease (33 months after the initial resection). A second patient, who had discontinued imatinib 6 months postoperatively relapsed 1 year after surgery and showed a clinical response to imatinib shortly after re-initiation of treatment.
Progression-free survival rates and actuarial survival were compared in all patients with response to imatinib (resected vs. not-resected) and calculated from the date of initiation of imatinib treatment (Fig. 1).
In 38 patients with either complete or partial response to imatinib who did not undergo resection of residual disease 1-, 2- and3-year progression-free survival rates were 94.4% (SE = ±3.8%), 77.4% (±7.6%) and 53.3% (±11.7%) compared to 100%, 92% (±8%) and 76% (±15%) in the resected group (n = 12). Calculated 1-, 2- and 3-year overall survival rates were 100%, 86.5% (±6.2%) and 81.1% (±7.7%) for the unresected and 100% for the resected patients with a median follow-up since initiation of imatinib of 25.7 (unresected) and 29.8 (resected) months, respectively.
GIST are known to be associated with a poor long-term prognosis. After complete resection of localized, non-metastatic GIST, the 5-year disease-specific or overall survival rate has been reported to be as low as 40–54%.6, 15, 16 Until recently, the outcome of patients with metastatic, unresectable or relapsed GIST has been even worse with median survival times ranging between 10–20 months.6, 15
Due to the location and the natural course of GIST, resection of metastatic disease has been rarely feasible (Table IV). In the pre-imatinib era, long-term survival has been observed only in a subset of patients after complete resection of metastatic disease. In the series of GIST from MSKCC,6 no long-term survivor was observed among a group of 18 patients with metastatic disease of whom 4 patients had undergone a complete resection of metastatic lesions. Ng et al.15, 16 reported on 19 patients with resectable peritoneal metastases of whom about 18% of patients were alive after 5 years. In a subgroup of 5 patients with completely resected liver metastases, one patient was still alive after 5 years. These figures seem to correlate with those of a larger series reported by DeMatteo et al.17 Among 34 patients who underwent resection of liver metastases from GIST and gastrointestinal leiomyosarcomas (GI-LMS), the 5-year survival rate after complete resection was 10%. In the only other large report of patients with liver metastases (±extrahepatic metastases) from GI-LMS (presumably GIST) reported by Lang et al.,18 3 of 15 patients (20%) in whom a complete resection was achieved were found to be alive after 5 years. Two of these three patients remained free of disease at the last follow-up. With the experience of resection of pulmonary metastases in sarcomas other than GIST19 these data may serve as a rationale to further evaluate aggressive surgical treatments in metastasized GIST.
Table IV. Survival Data of Patients with Localized and Metastasized GIST from Previous Studies
Median overall survival (months)
5 years (%)
10 years (%)
2Overall survival CR, complete resection; IR, incomplete resection; LR, local recurrence; MetR, metastatic recurrence; CR+TR, complete resection with tumor rupture; PM, peritoneal metastases; EHT, extrahepatic tumor.
Imatinib is the first effective systemic treatment in GIST and could therefore facilitate surgical approaches in patients with metastatic disease. The few previous reports regarding resection of metastatic disease in GIST are heterogeneous and vary upon localization of metastases, different populations and surgical standards by which the comparability toward recent studies is limited.
In our report, 12 of 90 patients with metastatic disease were considered for surgery and in 11 of these patients a complete resection was achieved (13%). Most patients in our series would have not been eligible for resective surgery because 10 of 12 patients had undergone tumor-related major surgery before; metastases were merely localized in the liver (8 of 12); and more than one metastatic site was found in 3 patients. In addition, most patients had multiple (median = 7.5) and large (median size = 5.4 cm) metastases before treatment with imatinib, suggesting that imatinib might improve the resectability of GIST patients with extensive, metastatic disease. Recently, Scaife et al.20 reported a complete resection rate of 12% after a 10–12 month treatment with imatinib among 126 patients (16 of 126) with primarily unresectable GIST that is consistent with our findings. Resection rates may vary considerably, however, for certain subgroups of patients. In 17 patients with isolated hepatic metastases, 6 were completely resected after treatment with imatinib resulting in a resection rate of 35% in our study. In addition, resection rates in our study varied considerably between the different participating centers (Essen, 20%; Hamburg,7%; Tübingen, 6%) that might reflect the different approaches for assessing resectability. Future prospective studies in this group of patients should therefore develop guidelines for assessing resectability. In this context, preoperative 3-dimensional reconstruction of hepatic vascular anatomy and volumetric assessment may improve the risk assessment in patients with borderline resectability.21
Data about the timing of resective surgery after imatinib treatment is scarce. Surgical strategies presently employed include resection after maximum response, or at the time of first progression. Hohenberger et al.22 described recently a single-center experience of resections of both responding and progressing GIST. In this series, the R0 resection rate in patients with potential resectability was 89% in patients with a partial response to imatinib (n = 9), 75% in patients who displayed secondary progression after a prior response to imatinib (n = 16); and only 21% in patients with primary disease progression (n = 14). These figures correlate very well with the resection rate of 92% in patients with a partial remission in our series and are therefore supportive of an approach performing a resection at the time of best response.
In our series, the rate of complications due to surgery was similar compared to the earlier report by Lang et al.18 (27 vs. 29%). Complications were mostly mild, however, and full recovery from surgery was usually observed within a few weeks. In contrast to most neoadjuvant treatments comprising “classical” chemotherapy, imatinib usually improves the performance status of patients with GIST.23 In our series all patients displayed a Karnofsky index of 90–100% at the time of surgery that might explain, in part, the good tolerability of surgery. In addition, the tumor load was actually decreased significantly after treatment with imatinib as the median number of metastases was reduced from a median of 7.5 to 2. One patient, however, had a prolonged deterioration of performance status due to symptoms consistent with short bowel syndrome with full resolution of symptoms only 8 months after surgery. It should be mentioned that this patient had a history of 3 tumor-related laparotomies including sigmoid colon resection and 2 independent resections of small bowel segments. With surgical death rates of 4–6% in previous studies,17, 18 the small number of patients in our study probably does not allow for any definitive conclusions to be drawn on surgical mortality. In concordance with recent reports,21, 24 however, resection of residual disease after response to imatinib is feasible whereas surgical treatments in patients with primary resistance to imatinib seem to be associated with increased mortality (8%) and postoperative morbidity.22
GIST responding to imatinib often show dramatic shrinkage and a complete loss of detectable glucose utilization in the 18FDG-PET scan that was also observed in our patients. In this series of patients hepatic metastases seemed to be completely cystic with liquified centers and capsule-like borders (Fig. 5) and radiological (CT) and functional (PET) imaging studies suggested complete necrosis in most cases. The predominant area of most resected tumors all showed signs of hyalinization or myxoid degradation albeit in all but one patient in whom residual tumor cells could still be found. In addition, a significant proliferative activity was observed in the pathological specimens of 3 patients despite the lack of any clinical evidence of tumoral activity such as increased FDG uptake or changes in the size of the tumor. These findings are interesting in several ways.
Even in patients favorably responding to imatinib, complete pathological remissions seem to be very unlikely. Hyalinization was observed in most areas of all tumor manifestations but a subset of cells obviously remained in a quiescent state of uncertain malignant potential in most patients as it has been observed in a case report published recently.24 These findings might explain the increasing number of patients who develop progressive lesions after a favorable initial response to imatinib25 and suggest that the effect of imatinib on these residual GIST cells is cytostatic rather than cytotoxic. Successfully depleting those cells might therefore improve the duration of remission or even the overall survival. Synergisms of imatinib with other protein kinase inhibitors, such as SU011248, RAD001 or PKC412, are presently investigated in clinical trials.26
Another interesting observation was the fact that almost no necrosis was detected in any of the resected specimens instead of vast areas of hyalinization, which suggests that the targeted inhibition of KIT mainly induces apoptotic pathways. Duensing et al.27 reported recently that in vitro KIT inhibition strongly affects the AKT/MTOR survival pathway resulting in decreased proliferation and increased apoptosis. Evaluating these residual cells for genomic changes or differences in the protein expression might therefore be helpful to elucidate mechanisms of resistance. Snap frozen tissue or even fresh tumor for short term cultures need to be obtained that might be difficult due to the very uneven and sometimes scarcely distributed residual cells in these specimens.
CT and PET or PET/CT might serve as a valuable tool to predict response to treatment with imatinib,28 although these diagnostic modalities do not seem to allow any prediction of histological response, and therefore all residual manifestations of disease should be suspected to harbor residual tumor cells.
The adjuvant treatment with imatinib in patients with GIST has not been proven to be beneficial and to date, patients should not receive treatment outside of controlled trials. Albeit most patients in our series have no evidence of disease after resective surgery, continuous imatinib treatment would hardly fall into the same category. In this series, we empirically decided to continue treatment with imatinib in all patients for at least 1 year. Eight patients were continued beyond 1 year, 2 patients stopped treatment after 6 months of whom one was found with a relapse after 6 months. Two patients refused a postoperative treatment. Blay et al.29 presented the data of a randomized trial evaluating continuous vs. intermittent use of imatinib in metastatic GIST. Within the group of patients who were discontinued after 1 year of imatinib and who had no evidence of disease (after either imatinib alone or additional surgical treatment), progression-free survival was similar compared to patients who had residual disease. Contrary to this, patients in the same clinical setting who had been treated with imatinib continuously showed a better progression-free survival. In the light of these findings, a continuous treatment with imatinib may apply best for patients in the clinical setting described in our study.
With a median follow-up of 2.5 years, the analysis of progression-free rate and overall survival suggests an advantage for patients who underwent resective surgery after treatment with imatinib compared to patients who responded but were not resected afterward. The differences observed in this analysis were not statistically significant and given the retrospective character of the analysis, any further conclusions cannot be drawn. Long-term data about the use of imatinib, in either an adjuvant or neoadjuvant setting, will need several years to mature and, to our knowledge, no prospective trial investigates the use of imatinib in patients with extensive metastatic disease. This imposes a considerable challenge on how to best manage these patients at present and our study may generate a hypothesis for a prospective trial in this group of patients.
In conclusion, our study shows that resection of residual disease after pre-treatment with imatinib in patients with metastatic GIST is feasible even in patients with advanced hepatic and peritoneal metastases. The observation of residual tumor cells in all but one resected specimen of patients with a favorable clinical response might serve as a rationale for an aggressive surgical approach but prospective trials are warranted. The prognostic impact of a surgical completion of an induction treatment with imatinib remains unknown and, therefore, the decision toward resective surgery should only be considered if minimal morbidity can be expected from the procedure.
Future prospective trials should possibly focus on the duration of preoperative imatinib, assessment of resectability, timing and extent of surgery as well as the role of post-operative imatinib in pre-treated patients.