GIST was mistaken for schwannomas, leiomyomas or leiomyosarcomas until the introduction of ultrastructural, immunohistochemical, and molecular biological techniques, which uncovered that GIST originated from myenteric nervous system[31], and ICCs were further suggested to be the cells of origin[32]. Subsequently, gain-of-function mutations in the tyrosine kinase receptor KIT and platelet-derived growth factor receptor-α (PDGFRA) were groundbreakingly found as the main oncogenic driver in GIST[33, 34], which encouraged the development of GIST targeted therapies[5]. In recent researches, Etwenty-six (ETS) variant 1 (ETV1) was also reported to overexpress in GIST and enhance the expression of KIT when binding target genes[1, 35]. However, nearly 10% to 15% of adult GIST and 85% of pediatric GIST are negative for KIT and PDGFRA mutations, as called wild-type (WT) GIST, which is a component of the Carney-Stratakis syndrome caused by the succinyl dehydrogenase (SDH)-mutations[36]. Whereas, liver was the most common site that GIST metastatic to, both in WT GIST and non-WT GIST, and LIM of GIST was always suggested to be related to the poor prognosis[37].
Nomogram, a quantitative tool for assessing risk and benefit, has been widespread applied in the medical domain for clinical decision-making[38]. In the previous studies, several nomograms have been developed and validated to predict the OS, cancer-specific survival (CSS) of GIST, recurrence-free survival (RFS) and disease-free survival (DFS) after surgical resection of GIST[25, 39-41]. However, the nomogram for predicting LIM of GIST has not yet been reported.
In the present research, we ultimately obtained 3797 GIST cases from SEER database and the LIM nomogram was further established based on five significant patient features, namely sex, tumor location, tumor size, N stage and mitotic rate. The discrimination performance, as demonstrated in the calibration plots (Fig.2b, c), was assessed by an internal bootstrap resampling approach. Furthermore, as shown in the ROC curves (Fig.3a, b), the LIM nomogram displayed better diagnostic efficiencies when compared with the other single variables, and the AUCs of the nomogram were 0.794 (0.778 to 0.808) and 0.775 (0.748 to 0.802) in the training and testing groups, respectively.
Although the classification, line(s) of differentiation, prognostication have long been the confusion and controversy of GIST, tumor size and mitotic rate were the widely accepted risk factors. By means of tumor size and mitotic rate, the first risk classification of GIST, constructed by Fletcher et al.[42], divided GIST patients into four sets, including very low risk, low risk, intermediate risk, and high risk. Previous studies indicated that the size and mitotic rate of GIST were proportional to poor prognosis[25, 39, 40], which corroborated our results. However, metastatic measures were not uncommon for postoperative GISTs, even that with small tumor size and low mitotic rate[43]. The results obtained from the multivariate analysis of Yang et al.’s study[11] presented that GIST size was not a significant prognostic factor of the liver metastatic GIST. In contrast, Mietinenn et al. suggested tumor size was the metastatic risk of GIST, rather than mitotic rate[44]. In our opinions, GIST with larger tumor size or higher mitotic rate tend to subject to earlier adjuvant therapy and more extensive resections, which may be a reason for the discrepancy in the previous studies[11, 45, 46]. Additionally, male GIST patients were more likely to have LIM than female ones, which was in accordance with the previous studies that indicated GIST had a mild male predominance[11, 47], however, the underlying mechanism remains unclear.
Recently, controversy exists surrounding the relationship between primary tumor location and prognosis of GIST. It is a common dogma that gastric GIST (G-GIST) has a more favorable behavior when compared with small intestinal GIST (SI-GIST)[17]. Together with tumor size and mitotic rate, Miettinnen et al. added tumor location as a poor prognostic factor for the construction of Armed Forces Institute of Pathology (AFIP) classification[42, 48]. Based on approximately 2000 cases, SI-GIST resulted in a relatively higher risk of metastasis and tumor-related death, particularly with the tumor size exceeding 5 cm[48]. Anatomic site was also reported to be the significant independent predictor of OS[49], CSS[50] and RFS[51], with SI-GIST accompanied with significant disadvantage in the prognosis as compared to G-GIST. Furthermore, Kukar et al. found that younger patients with SI-GIST had a tendency to be presented with distant metastatic disease and larger tumor size[50]. In addition, the proportion of KIT exon 9 mutation was strikingly higher in SI-GIST than that in G-GIST, which may be the explanation of poorer prognosis of SI-GIST[52]. Inversely, several studies based on SEER database revealed comparable prognosis between small bowel and gastric GIST. After adjusting the confounding variables on a population based level, Guller et al. found that SI-GIST and G-GIST shared similar OS and CSS, which was contrary to common belief [17]. These results reflect those of Giuliano et al.[53] who further found that, although SI-GIST did have more aggressive features, SI-GIST patients were also more likely to undergo surgery than G-GIST (89.8% SI-GIST vs. 78.7% G-GIST), leading to the comparable survival outcomes. However, the previous studies always investigated SI-GIST as an entire cohort. As displayed in Fig. 2a, G-GIST seemed to share similar score with duodenal GIST (D-GIST), while jejunal GIST (J-GIST) patients were more likely to have LIM than ileal GIST (I-GIST) patients. Although I-GIST and J-GIST were reported to share compared prognosis in the study of Feng et al[54], we suggested that more aggressive treatment should be taken into consideration to J-GIST for the high risk of LIM. In comparison with colon GIST (C-GIST), rectal GIST (R-GIST) tended to have more positive prognosis in spite of the less likelihood of surgical resection[50], analogous result could be found in the current study. Moreover, the aggressive course of extra-gastrointestinal GIST (EGIST) was suggested to be akin to SI-GIST[55], whereas according to the present LIM nomogram, EGIST seemed to shared similar prognosis with D-GIST, irrespective of I-GIST and J-GIST. Sample size and potential bias may result in such differences.
Unlike other solid tumors, lymph nodal involvement is extremely rare in GIST patients and lymph node dissection is not routinely suggested during the surgical treatment[56, 57]. However, in the present study, although the rate of lymph node metastasis (LNM) in the entire cohort was low (3.3%), LNM was significantly associated with LIM. This finding was consistent with that of Gaitanidis et al. who also found LNM was an independent prognostic factor of worse overall survival in patients with metastatic GIST[58], which further revealed that the evaluation of regional lymph nodes could be taken into consideration when undergoing surgical resection in patients with metastatic GIST. In addition, Li et al. found that lymphadenectomy was associated with an risk of mortality in GIST patients, which may be attributed to the destroy of the immune micro-environment in the normal lymph nodes and increasing postoperative morbidity and mortality caused by surgical trauma[59, 60]. What’s more, GIST patients with SDH complex deficiencies tended to have LNM. The SDH-deficient related disease like WT GIST (a component of the Carney-Stratakis syndrome) was reported to have high rates of LNM (29%)[61], hence the resection of enlarged nodes in SDH-deficient neoplasms was recommended in the National Comprehensive Cancer Network (NCCN) guidelines[62].
However, the current study is subject to several limitations. The study is a retrospective analysis, systematic and prospective data were lacked. External validation at other institutions was also lacked in our research, which may lead LIM nomogram to be overfitting. In addition, several critical clinicopathologic variables were required, especially the administration of tyrosine kinase inhibitors. If the information of co-morbidities, immunohistochemistry, and other laboratory values could be available for the construction of LIM nomogram, the results of our study might provide more valuable therapeutic measures for clinician.