The tumor-stromal ratio as a strong prognosticator for advanced gastric cancer patients: proposal of a new TSNM staging system

Recent figures show that gastric cancer (GC) is now the second commonest cause of cancer related mortality in China with over 498,000 deaths in 2015 [1]. Despite significant advances in GC early diagnosis, staging system, and treatment, the 5-year overall survival (OS) of GC remains around 40% [2, 3]. One of the key causes is that GC is a heterogeneous disease in its biologic behaviors, but current prognostic and predictive factors for GC do not provide optimal risk stratification, thus additional information is necessary to improve tailored treatment for the individual patient [4, 5]. How to establish clinically relevant subtypes that would encompass this heterogeneity and provide useful clinical information has been considered [6]. Although many new and modified subtypes have been proposed, a tumor-based TNM staging system is still the dominant instrument to guide treatment strategy for GC patients, and insufficient attention is paid to the underlying tumor microenvironment (TME) evolution, which is the underlying cause for tumor heterogeneity, divergent cancer aggressiveness, and clinical outcomes [7, 8].

The TME concept considers the interplay between cancer cells and stromal cells as the primary driving force for cancer progression. In the past, tumor cells have been the main target for therapeutic interventions [9]. Nowadays, however, accumulating evidence illustrates that tumor stromal, as an integral component of TME, could facilitate tumor progression [10]. Some essential components of stroma, including tumor associated macrophages (TAM), cancer associated fibroblasts (CAF), and tumor-infiltrating lymphocytes (TIL) [11‐13], have been evaluated. Moreover, some critical events have also been investigated and applied in clinical setting, such as targeting the immune cells and angiogenesis [14, 15]. The novel concept that the TME is built through rate-limiting steps during multistage carcinogenesis has been established. Thus, TME represents a unique tumor property and a robust target for cancer treatment and prevention [16]. However, considering the complexity components of TME, overall consideration about the evolution of stromal should be highlighted [17, 18].

The morphological evaluation of the ratio of stromal at the most invasive part of primary tumor (tumor-stromal ratio, TSR) is gaining momentum as evidence strengthens for the clinical relevance of this TME parameter. TSR was first reported as an independent stromal parameter for survival as compared with lymph node status and tumor stage in colon cancer, and there was a preliminary attempt to adding the stromal parameter to the ASCO colon risk criteria [19]. It was concluded that TSR should be integrated into the standard pathology reports in addition to the current TNM classification [20]. As a newly identified prognostic factor, it was also found to be significantly associated with prognosis of esophageal squamous cell carcinoma, breast cancer, and cervical cancer [21‐23]. Accumulating evidence suggests that the important role of TME could be generally assessed by the TSR evaluation of simple hematoxylin and eosin (H&E)-stained tumor sections [19].

With these considerations in mind, we hypothesize that TSR would be a suitable TME parameter for GC that may parallel the evolving tumor-stromal interactions. We have undertaken a comprehensive analysis of the TME of GC, and evaluated the prognostic and predictive role of TSR in GC, and stratify the GC patients into four groups based on TNM staging system and TSR. TSR could be easily implemented in routine daily pathology diagnostics, as it is simple to determine, reproducible, and performed in a short time.

We have shown that TSR is a useful TME criterion for the prediction of tumor survival in patients with GC. Herein, we answered three novel questions. First, do we observe a suitable TME parameter that parallels the evolving tumor-stroma interactions? Second, is the TME parameter comparable or more powerful than the cancer cell-based histopathological criteria in predicting prognosis? And third, in which way can the TME parameter be integrated into standard pathology reports in addition to the current TNM staging system for better risk classification?

Compared to the progress in TME theory and TME-based therapy, the application of the TME concept in cancer prognosis is relatively stagnant [26, 27]. Although researchers have shown the predictive power arises from genes expressed by stromal cells [28], identifying of new convenient TME prognostic factors and integrating into the current TNM staging system may have profound influence for tumor prognosis and therapeutic strategies. Alternatively, morphometrical analysis enables accurate quantification of various tissue components when compared with qualitative systems. Here, we evaluated the TSR by pathologists. Our results indicated that TSR was a new and strong independent prognostic factor for GC. TSR was correlated with TNM stage (P = 0.031) and the risk of death was increasing 91% for patients in the stromal-high group compared to patients in the stromal-low group [HR = 1.911 (95% CI 1.427–2.559), P < 0.0001]. Receiver operating characteristic curve showed that TSR performed well in predicting the clinical outcomes of GC patients compared to other factors [area under the curve: 0.615 (95% CI 0.584–0.665), P < 0.001] (Fig. 2).

TSR is a robust prognostic factor in cancer, while it may be argued that to what extent the TSR could represent the stromal status. Based on the progression of TME theory, we hypothesize that evolving tumor-stromal interactions contributed to the presence and variable of TSR. Previously, we suggested that tumor cells can activate their adjacent stroma during the TME evolution process. The activated stromal, which receive and elicit oncogenic signals and undergo quantitative and morphological changes, was the source of promoting cancer progression [18, 29]. There are several theories as to why a high proportion of stromal within a tumor may infer a poor prognosis. First, it has been hypothesized that tumors with a greater proportion of stroma are able to produce more stroma-derived growth factors thus increasing the overall tumor burden [30]. Second, it has been suggested that the relative amount of desmoplastic fibrosis may have a role in reducing the accessibility of tumors to the immune response by encapsulating the malignant cells and preventing their destruction [31, 32]. More importantly, high proportion of stromal means that more stromal are activated by fewer cancer cells, indicating that the coevolution between cancer cells and tumor stromal is more effective, and consequently, results in poor overall survival. Our result showed that the likelihood of tumor invasion was increased significantly for patients in stromal-high subgroup compared to those in stromal-low group (P = 0.011). Our study underlined this hypothesis that an increased amount of stromal involvement, even when it is detected in only a small part of the total tumor mass, can be linked to an unfavorable prognosis independent of other prognostic parameters. Possibly, this particular part of the tumor has obtained the capability to orchestrate its direct environment to facilitate its invasive and metastatic behaviors [33, 34]. Thus, we answered the first and second novel questions that TSR would be a rational and typical TME parameter for tumor prognosis.

In addition, we integrated the TSR into TNM staging system, and proposed a new staging system creatively. This new staging system based on the pathobilogical behavior for GC is the TSNM (tumor-stromal node metastasis) system (Fig. 2), which integrates cancer cell status (invasive depth, lymph node spread, distant metastasis) and stromal behaviors (activated, inactivated). This study highlighted the importance of integrating cancer status and stromal status. Deep analysis revealed the better predictive performance of the TSNM (Fig. 2). This new staging system is a good paradigm to encompass the tumor heterogeneity and provide useful clinical information can improve current high-risk stratification methods.

Since we have shown the prognostication power, discrimination ability and availability of TSR and TSNM staging system for GC prognosis in this study, validation of convenience and reproducibility of TSR and TSNM should be performed for further application. The determination of TSR proved to be a quick and convenient procedure, which can be performed during pathological examination, using standard H&E sections without the need for additional costs above standard diagnostics. It is a simple, relatively inexpensive morphometrical measurement. To test the reproducibility, it should be emphasized that if the primary purpose is to find an approach for daily practice, the impact on daily routine should be minimal without a significant increase in the pathologist’s time [11]. Recent studies have shown that the Cohen’s kappa coefficient is high in some cancer patients, including those with breast cancer [23, 35], esophageal cancer [22], cervical cancer [21], and colorectal cancer [19, 20, 36]. These kappa values can be translated as almost in perfect agreement when evaluating TSR. Therefore, we believe that the TSR may be a candidate parameter to be implemented as a standard procedure within routine pathology laboratories. Although the results are promising, the limitations should be realized. At present, the criteria for unequivocal identification of TSR are currently unclear, and there are no established thresholds. More validation studies should be performed before the parameter can be integrated into future clinical trials, translational research, and diagnostic practice. Then, sufficient evidence would facilitate the widespread use of TSR and the TSNM staging system.

In the study of Ueno and colleagues, stromal was classified into three types based on the desmoplastic reaction (DR) resulting from stromal fibroblasts. Prognostic role of stromal type has been established [37, 38]. Given the complexity of TME, the stromal types could be defined based on the stromal cell ratio, stromal cell density, ECM density, and other parameters. However, no evidence has shown which definition of stromal types is best. In the future, we would classify stromal type deeply based on stromal cellular components (macrophages, cancer associated fibroblasts, tumor-infiltrating lymphocytes, etc.) and non-cellular components (ECM, etc.). In addition, only two patients (0.4%) with T3 tumors were included in this study such bias of distribution in populations would influence the result. In the future, multiple-institution based validation should be performed to verify the proportion of T3 tumors.

In conclusion, TSR is a convenient and useful TME parameter for oncologists to obtain more prognostic information, and the concept of integrating the TME parameter into the TNM staging system is available. TSR has shown to be a prognostic parameter and is easy to determine, is reproducible, and quickly performed for all types of GC. Because of its low cost, availability, convenience, high reproducibility, and importance in prognosis, TSR is a candidate TME parameter that could easily be implemented in routine pathology diagnostics, and TSNM staging system has been established to optimize risk stratification for GC. The value of TSNM staging system should be validated in further perspective study.