In the previously discussed studies, at least half, if not the majority, of patients had GIST of stomach origin. At the more rare sites, esophagus, duodenum, and rectum, the ability to achieve complete resection is made more challenging by anatomic constraints, resulting in potentially morbid operations. In the esophagus, esophagectomy is frequently needed, which requires entry into the abdomen, chest, and in most cases, the neck. In the duodenum, depending on the tumor location and the extent of involvement, pancreaticoduodenectomy (Whipple procedure) may be needed. In the rectum, tumors may require abdominoperineal resection, which removes the anal sphincter and involves the creation of a permanent colostomy. For GIST at these specific anatomic sites, neoadjuvant imatinib is an attractive strategy to enable complete and potentially less morbid resection (e.g., sphincter-preserving excision in the rectum).

For esophageal GIST, case reports of neoadjuvant imatinib have been published [13–15]. In the majority of these studies and depending upon the author’s clinical judgment, patients still required esophagectomy. However, the benefit of neoadjuvant treatment in these studies seemed to be in converting patients from unresectable to resectable and preventing tumor rupture for large, bulky tumors.

Marano et al. recently reviewed the literature for duodenal GIST [16]. These tumors actually appear to have a better overall prognosis compared to GIST at other gastrointestinal tract sites, including the stomach. The authors point out that duodenal GIST, despite their location, typically displace rather than invade surrounding structures. Tumors tend to grow opposite the duodenal lumen and toward the abdominal cavity. In support of this concept, Colombo et al. reported data from a multi-institutional cohort of patients with duodenal GIST and found that the type of resection (conservative duodenal resection versus Whipple) does not impact the clinical outcome [17]. In this series, 11 patients actually received neoadjuvant imatinib (400 mg daily) for a median of 8 months. Similar to other retrospective data, nine of these patients (80 %) demonstrated objective response. In 6 of the 11 patients (55 %), pancreaticoduodenectomy was avoided and conservative resection was feasible.

For rectal GIST, several case reports and small case series have shown that neoadjuvant GIST can indeed downsize tumors to allow for sphincter-preserving surgery. Jakob et al. described 39 patients with rectal GIST of which 16 received neoadjuvant imatinib [18]. The authors found a higher rate of margin-negative complete resection in those who received neoadjuvant therapy compared to those who did not. Tielen et al. also had similar findings with 12 rectal GIST patients who received neoadjuvant imatinib, although five patients still required abdominoperineal resection and two patients required posterior exenteration, an even more extensive operation [8].

In part as a result of some of the data discussed, consensus guidelines do recognize the potential value of neoadjuvant imatinib in the management of locally advanced GIST. The European Society of Medical Oncology (2010) recommends preoperative imatinib as a treatment option if this results in “less mutilating surgery and lower risk of tumor bleeding/rupture” [19]. The National Comprehensive Cancer Network (NCCN, STS Guidelines 1.2015) recommends neoadjuvant imatinib in patients with resectable GIST but with the risk of significant morbidity. This includes patients who may require multivisceral resection or abdominoperineal resection due to locally advanced disease. Interestingly, these guidelines also suggest that if these patients have disease progression, surgery should be considered for salvage, if feasible.

It is important to note that the tumor response to imatinib is rarely complete. By cross-sectional imaging, imatinib rarely results in complete tumor regression with no measurable disease. Similarly, under the microscope, histologic response to imatinib rarely induces complete necrosis without any viable tumor cells. In fact, tumor response by histology is frequently quite variable even at different components within the same tumor, ranging from 10 to 90 % reduction in tumor cellularity [20, 21].

The presence of any residual, viable tumor cells after prolonged imatinib treatment implies the possible risk of development of resistant clones. This was highlighted in a case report by Haller et al. [22]. This patient had initially unresectable GIST, received 10 months of imatinib (400 mg daily) with tumor response followed by surgical resection. Detailed histologic examination of the resected tumor specimen identified multiple remnant tumor microfoci, each measuring less than 0.3 cm. Surprisingly, in comparison to the KIT mutation status on the pre-imatinib biopsy, analysis of the resected tumor specimen identified the additional new point mutations.

The potential detrimental effect of prolonged neoadjuvant imatinib was shown recently in a study by Bednarski et al. [23] This retrospective review of patients included 41 patients with locally advanced, primary GIST who had undergone preoperative imatinib for a median duration of 315 days (10.5 months) prior to surgical resection. In this group of patients, neoadjuvant therapy greater than 365 days (12 months) was associated with a higher risk of recurrence.

Interestingly, GIST treated with imatinib can also exhibit a variety of other histologic changes. Although GISTs are more commonly spindle shaped, treated tumors may develop a purely epithelioid morphology and even a tubulopapillary growth pattern [24]. The expression of CD117 may also be reduced or even lost after treatment, a characteristic found to be associated with disease recurrence by Mearadji et al. These CD117-negative GISTs include high-grade, anaplastic sarcomas, which have been observed in both imatinib-treated and treatment-naïve tumors [25]. In rare cases, imatinib treatment may also result in the development of other histologic lineages within the tumor, including rhabdomyoblastic, cartilaginous, and osseous transdifferentiation [26, 27]. The clinical significance of transdifferentiation in imatinib-treated GIST is currently unknown.

Complete surgical resection is the goal of treatment for patients with localized primary GIST. In the subset of patients with locally advanced, borderline resectable or unresectable tumors, the ability of imatinib to induce measurable tumor shrinkage may offer the renewed opportunity to achieve complete resection. In support of this, several retrospective studies seem to suggest a high rate of objective response (70–80 %), allowing for resection in these patients and in many cases less morbid and function-sparing surgery (e.g., rectal GIST). Prospective studies such as the RTOG 0132/ACRIN 6665 also support the use of neoadjuvant imatinib, however, the data are not as robust. The results of the CST1571-BDE43 or APOLLON trial will hopefully provide more data to further assess the efficacy of neoadjuvant imatinib.

Several issues arise regarding the use of neoadjuvant imatinib in locally advanced, primary GIST. First, selection of appropriate candidates should be better defined. An important component of this decision should be tumor genotyping, including mutation status. Tumors with a KIT exon 9 mutation will be more sensitive to a higher dose of imatinib (800 mg) [28]; tumors with a PDGFR-alpha mutation in D842V (exon 18) will likely be resistant to imatinib [29, 30]. Second, a key question is the appropriate duration of imatinib therapy to capture the window of opportunity between maximal tumor response and disease progression or development of resistance. Gold and DeMatteo proposed surgery within 6 months of therapy [31] and based on the data from Bednarski et al., this should certainly be not more than 12 months [23]. We advocate that in all cases, this decision to discontinue neoadjuvant treatment for surgery should be personalized to the individual patient and made in the setting of a multidisciplinary discussion (e.g., radiologist, medical oncologist, surgical oncologist). Third, adequate biomarkers of response are lacking. Metabolic response, as measured by decrease in FDG avidity by PET, has been explored by Goh et al. and Van den Abbeele et al. with somewhat conflicting results [32, 33]. We would favor other response biomarkers, including histologic and even molecular (e.g., GLUT4) [33]. Fourth, the efficacy of other targeted therapies (e.g., sunitinib, regorafenib) should be explored in the neoadjuvant setting, especially for patients who have tumors with evidence of imatinib resistance. We also support investigation of combination therapies and inclusion of novel therapies (e.g., immunotherapy) in the neoadjuvant setting.