Clear indications for resection include isolated lesions that are amenable to complete resection with expected negative margins [3]. Symptoms that represent traditional indications for intervention, such as hemorrhage refractory to endoscopic management, bowel perforation, and obstruction, also clearly mandate exploration [8]. Localized lesions larger than 5 cm in anatomically difficult locations may benefit from preoperative therapy, with each clinical scenario being assessed individually (see below regarding neoadjuvant therapy).

The management of small (<2 cm) incidentally discovered, asymptomatic GIST is less clear. These lesions are often found incidentally during upper endoscopy. Some are potentially resectable by endoscopic means, however, this approach requires specialized skills. The endoscopic management of GIST is covered in detail in a separate chapter in this book. Endoscopic ultrasound (EUS) is an excellent diagnostic adjunct to assess the size as well as the depth of the lesions. Formal resection is indicated in the presence of high-risk features on EUS (echogenic foci, ulceration, or irregular margins). Otherwise, serial (every 6–12 months) endoscopic surveillance may be recommended for small, low-risk lesions, especially in a patient whose comorbid conditions might be otherwise prohibitive [3, 4].

Complete resection is the treatment of choice for GIST, either as primary therapy or after a favorable response to neoadjuvant therapy. While the optimal margin of resection is controversial, the decided trend is toward less aggressive resections. This approach is grounded in a tumor physiology lacking significant intramural spread or areas of skip metastasis, as well as favorable outcome data for these more limited resections. In addition, focal, microscopically involved margins (focal R1 resections) do not adversely impact survival on long-term follow-up [1]. Furthermore, the resection of GIST does not mandate the routine removal of large lymph node basins, similar to most soft tissue sarcomas lacking a predilection for lymph node invasion. Lymph nodes should only be removed if involvement is noted on preoperative imaging or during intraoperative exploration [2].

Patients with early disease who are clearly amenable to resection are best served by resection. However, for patients deemed unresectable, the use of imatinib (Gleevac®, tyrosine kinase inhibitor) may be indicated as first-line therapy. A favorable response to therapy can lead to downsizing and subsequent resection. For those patients with unfavorable (large, difficult locations) yet clearly resectable tumors, the decision to use neoadjuvant imatinib is less clear. The key issue, yet to be determined, is the threshold at which the benefit of the therapy outweighs its risks (albeit minor) and costs. Specific biology can be critical; for example, tumors harboring exon 9 mutations appear to require higher doses of imatinib to attain a clinical response [9]. Of course, patients not harboring the c-kit mutation have no clear response to imatinib and should not be considered for neoadjuvant therapy with the available drugs. Indications for the preoperative use of imatinib are listed in Table 2. Although not a strict requirement if primary resection is chosen, a biopsy will be necessary if the decision is made to institute neoadjuvant therapy. In addition to reducing surgical morbidity, a response to therapy has also been shown to yield a reduced rate of tumor rupture during subsequent resection [10, 11].

When instituting targeted therapy in advance of a planned resection, there are two traditional approaches to determine the timing of operative intervention. The first strategy is to allow a set time after the institution of chemotherapy, typically 8–12 weeks based on the original RTOG 0132 protocol, before proceeding with surgical intervention [12]. This allows a reasonable period for therapy based on early phase II trials in advanced disease [13]. The second, more contemporary option is intervention based on assessing for a maximal response to therapy by serial imaging. For the majority of tumors that respond to imatinib, resection can be performed with maximal benefit of preoperative therapy, but well before the development of resistant clones [14]. The median interval to maximal tumor response has been shown at 28 weeks, with a plateau at 34 weeks [15]. The NCCN recommends treatment until a plateau response is reached, as evidenced by two consecutive stable imaging studies [3]. Our preference is to obtain baseline cross-sectional imaging (CT or MRI) before institution of neoadjuvant therapy, and serial repeat imaging until a plateau response is reached, at which time an operative intervention is performed.

Early observation and regular imaging are critical, as disease progression requires prompt reassessment [4]. Repeated PET, although highly accurate, is not always available or feasible for surveillance in advance of operative intervention [3, 16]. While CT is the workhorse for perioperative imaging, interpretation can be difficult. For example, an initial response pattern of GIST can manifest as swelling and be misinterpreted as tumor progression. Many criteria have been described to objectively quantify tumor response by imaging. RECIST criteria, while widely used and accurate in many tumors, can underestimate GIST response to therapy. The criteria by Choi et al. (>10 % decrease in tumor size and >15 % decrease in tumor density) more reliably predict therapeutic response by CT [17, 18].

The perioperative scheduling of neoadjuvant therapy varies significantly between imatinib and the newer targeted therapies. Imatinib can be stopped immediately prior to surgery and reinstituted as soon as oral medications are tolerated. Newer agents such as sunitinib and regorafenib are more systemically caustic and should be stopped a week prior to operative intervention. These second-line agents should be restarted more judiciously based on individual patient recovery [3].

Potential approaches for resection of GIST include endoscopic, open, laparoscopic, hand-assisted laparoscopic, robot-assisted, or a combination of these techniques. In achieving an R0 resection, application of these techniques can vary widely according to tumor size, location, and extent of local invasion. Traditional dogma has relegated the laparoscopic approach to small, easily accessible tumors over the fear of violating margins and tumor spillage. This concern is valid, such that great care and gentle handing of the specimen should be taken with any approach. A specimen retrieval bag should be utilized during every minimally invasive approach [3].

Contemporary evidence demonstrates equivalent oncological outcomes, along with the expected reduced morbidity conferred by the minimally invasive approach [7, 19–21]. One caveat of early data is short follow-up, small studies, and a significant selection bias toward smaller and generally more favorable tumors in the minimally invasive approach. Hand-assisted surgery can provide many of the advantages of open and minimally invasive techniques and until recently was explicitly recommended by the NCCN for tumors over 5 cm to help ensure safe specimen handling [3, 22]. Robot-assisted surgery can be particularly effective for resections in difficult locations, such as the pelvis. The only contraindication to the minimally invasive approach is the inability to safely conduct an oncologically sound procedure. The operative approach chosen should be tailored to the tumor characteristics as well as the level of comfort and experience of the surgeon.