We have proposed a combined analysis of integrated tumor stromal features as a useful strategy to evaluate cancer progression and patient survival in GC based on our studies focused on the co-evolution of tumor cells and tumor microenvironment and [19, 21, 22]. This study was designed to explore the feasibility of this combined strategy. In addition, an improved automation method to analyze the digitalized images was used to ensure both reproducibility and good performance in this study. As Fridman  suggested, such methods would pave the way to better understanding the complex tumor microenvironment, as well as to the routine evaluation of parameters for clinical management of cancer patients. Herein, 184 GC cases were included to evaluate the prognostic values of optimized conventional pathological prognostic factors, cellular molecular factors, immune factors and the combined features. This is the essential step towards establishing a workable prognostic system integrating both clinico-pathological, tumor and stromal features in our series studies [19, 21, 22].
Of 184 cases, the demographics and clinico-pathological characteristics are similar to those reported in other large series of GC population . Our results showed that the expression of MT1-MMP was frequently correlated with increased recurrence risk, but the difference in relapse location was not statistically significant. These results were similar to previous report . MT1-MMP plays important role in degrading types I and IV collagens to facilitate cancer cells spreading. In addition, MT1-MMP can promote angiogenesis and micrometastasis via vascular route .
With regard to immune cells, the nature, density and location are important parameters to comprehensively evaluate the in situ immune reaction and the specific role in cancer progression. In this study, CD11b + immunocytes were mostly located at the invasive front. The difference in CD11b + immunocytes density was statistically significant between lymph nodes metastasis and non-metastais subgroups. Furthermore, the CD11b + immunocytes density was higher in early than advanced GC patients, similar to the reports by Sconocchia et al. and Ladoire et al. [27, 28]. Hence, we hypothesized that CD11b + immunocytes could prevent the lymph nodes metastasis by active immunosurveillance process.
The prognostic value of traditional clinicopathological prognostic factors has been validated [29, 30]. Interestingly, some studies reported that the LNR was a better predictor of patient outcome than lymph nodes status only. LNR may be an alternative stratification in cases where few nodes are retrieved [3, 8]. LNR has also been adopted by the Japanese Gastric Cancer Association (JGCA) .
Researches focused on molecular factors for cancer progrosis have attracted increasing attention [32, 33]. In this study, MT1-MMP expression and CD11b + immunocytes density were independent prognostic factors, which partly validated others’ conclusions about MT1-MMP and CD11b + immunocytes. Kanazawa et al.  reported that MT1-MMP expression could be considered as a useful independent predictor of outcomes in colorectal cancer patients. The results presented by Mahmoud et al.  confirmed the presence of efficient immunologic antitumor defense mechanisms in human breast cancer. It is proposed that immune score would identify a population of patients who would derive substantial benefit from further stimulating their immune response . Several studies have also provided evidence of immune criteria to predict which tumors have a high risk of death [35, 37].
Given the fact that tumor biology is often dictated by several essential cellular and microenvironmental alterations, it may be naive to think that single factor would be enough as prognostic factors in cancer . Solutions are now being explored by analyzing multiple factors with tissue microarrays, which has been emerged as an essential tool in the discovery and validation of tissue biomarkers . To our knowledge, combined analysis is a promising method to translate experimental results into clinical application . Based on our results and current knowledge in cancer progression, we proposed a new prognostic model that combines pathological, cellular and molecular features. This study showed that the combined features were independent prognostic factors for OS. The death risk of GC patients in group II was increased by 200% and this combined features would better predict GC patients’ outcomes.
The development of tumor biomarkers ready for clinical use is complex, and a useful prognostic marker must be a proven independent, significant factor, that is easy to determine and interpret and has therapeutic impact . Although the combined features described herein could address these conditions, the promising results are based on retrospective analysis, which is the limitation of this study. Prospective randomized clinical trials to evaluate the clinical utility of a prognostic or predictive biomarker are the gold standard, but such trials are costly and difficult to implement, and more efficient indirect “retrospective analysis” using archived specimens would be an alternative method for a long time .