The notion that all GISTs, irrespective of tumor size and mitotic rate, have a quantifiable risk of progression and malignant potential is controversial. Based on both autopsy and population level studies, incidentally identified “small” or “micro” GISTs, defined as GISTs measuring less than 1 cm in size, are quite common, occurring in the stomachs of 23 % of patients older than 50 years of age [12], in 35 % of patients with gastric cancer [13], and in 10 % of patients with gastroesophageal cancer [14]. Pathological analysis has demonstrated that these small GISTs already exhibit KIT and PDGFRA mutations in ~70–90 % of cases, but these tumors are clinically indolent [15, 16]. There are two retrospective analyses of small submucosal GISTs resected endoscopically, and both studies demonstrate stable disease on follow-up. Bai et al. analyzed 25 patients with small/micro-GISTs (84 % very low risk, 12 % low risk, and 4 % intermediate risk) and demonstrated no recurrence or metastasis with a mean follow-up of almost 12 months [17]. Similarly, Catalano and colleagues demonstrated a 5-year disease-free survival rate of 100 % in 10 patients with small GISTs that were resected endoscopically [15].

In 2001, the NIH convened a GIST Workshop and established a NIH consensus classification system which risk stratifies GISTs based on tumor size and mitotic count (Table 2) [18]. Given the aforementioned progression risk of even small tumors with low mitotic counts (especially distal lesions with mixed pathologic features), this classification system does not dichotomize GISTs into benign and malignant lesions. Rather, it classifies GISTs into very low-, low-, intermediate-, and high-risk categories [18]. While the performance of this classification system has generally been demonstrated to be quite satisfactory [19–21], it is notable that the NIH classification schema does not take into consideration several other important predictors of aggressive tumor biology. The multiple series by Miettinen et al. mentioned above highlight the biologic differences of GISTs based on tumor location, with small bowel GISTs demonstrating a more aggressive biology than gastric GISTs of equal size. The Memorial Sloan Kettering Cancer Center (MSKCC) group demonstrated that a nomogram that includes tumor location (small bowel versus stomach) in addition to tumor size and mitotic count performed better than the NIH classification system in predicting recurrence-free survival (concordance probability 0.76 vs. 0.70, P = 0.04) [22].Based on MSKCC’s report, the National Comprehensive Cancer Network (NCCN) adopted a more comprehensive risk stratification schema incorporating tumor location, as depicted in Table 3 [23, 24]. Inclusion of tumor location allows for more accurate risk classification given that extra-gastric GISTs have a poorer prognosis when compared to gastric GISTs.

It should be noted that the size and mitotic count thresholds of 5 cm and of 5 mitoses per 50 HPFs, respectively, first employed by Miettinen and colleagues and later adopted by the NIH and NCCN in their classification systems, were arbitrary and based on expert consensus. In a retrospective analysis of 929 resected GISTs in the pre-imatinib era (1980–2000), Rossi et al. constructed a risk stratification system that included tumor size and mitotic rate as continuous variables as opposed to arbitrarily dichotomized categorical variables [25]. The resultant nomogram demonstrates that the added risk of a GIST with 6 mitoses/50 HPFs is not equivalent to that of a GIST with 12 mitoses/50 HPFs, as would be the assumption of the prior risk stratification schema that uses them as categorical variables (Fig. 1). Their model demonstrated improved discriminative ability (C-index 0.72) when compared to the NIH (C-index 0.64) and the NCCN (C-index 0.63) risk stratification systems, suggesting that future risk prediction models should be constructed with continuous variables for better discriminative performance.

These same authors also proposed complementing the clinicopathologic risk stratification with the tumor’s mutation profile [26]. Pathologic analysis of untreated GISTs demonstrates that tumors with KIT exon 9 and exon 11 mutations are biologically more aggressive than wild-type GISTs or those with PDGFRA mutations [27, 28].In studying 451 primary GISTs treated only with surgery and no neoadjuvant or adjuvant therapy, Rossi and colleagues corroborated these findings, prognostically classifying GISTs into three distinct groups with decreasingly favorable biology: Group I, including GISTs with mutations in PDGFRA exon 13, KIT exon 17, and BRAF (reference group); Group II, including GISTs with mutations in KIT exon 17, PDGFRA exon 18 D842V, and PDGFRA exon 14, as well as triple-negative GISTs (HR 3.06, 95 % C.I. 1.09–8.58); and Group III, including GISTs with mutations in KIT exons 9 and 11 and PDGFRA exon 18 other than D842V (HR 4.52, 95 % C.I. 1.65–12.37) [26]. However, in current clinical practice, resected GISTs are not routinely submitted for mutational analysis, nor is it recommended by the NCCN GIST Task Force [24], as more data are needed to justify its added value in prognostication, taking into consideration the added cost and expertise required to do it.

It should also be noted that symptomatic presentation [29], tumor rupture [30, 31], cellularity, ulceration, and mucosal invasion [32] have all been shown to be important prognostic factors, but they are less commonly used for risk stratification.