Significance of Tumor–Stroma Ratio (TSR) in Predicting Outcomes of Malignant Tumors

Background and Objectives: The present study aimed to elucidate the distribution and the prognostic implications of tumor–stroma ratio (TSR) in various malignant tumors through a meta-analysis. Materials and Methods: This meta-analysis included 51 eligible studies with information for overall survival (OS) or disease-free survival (DFS), according to TSR. In addition, subgroup analysis was performed based on criteria for high TSR. Results: The estimated rate of high TSR was 0.605 (95% confidence interval (CI) 0.565–0.644) in overall malignant tumors. The rates of high TSR ranged from 0.276 to 0.865. The highest rate of high TSR was found in endometrial cancer (0.865, 95% CI 0.827–0.895). The estimated high TSR rates of colorectal, esophageal, and stomach cancers were 0.622, 0.529, and 0.448, respectively. In overall cases, patients with high TSR had better OS and DFS than those with low TSR (hazard ratio (HR) 0.631, 95% CI 0.542–0.734, and HR 0.564, 95% CI 0.0.476–0.669, respectively). Significant correlations with OS were found in the breast, cervical, colorectal, esophagus, head and neck, ovary, stomach, and urinary tract cancers. In addition, there were significant correlations of DFS in breast, cervical, colorectal, esophageal, larynx, lung, and stomach cancers. In endometrial cancers, high TSR was significantly correlated with worse OS and DFS. Conclusions: The rate of high TSR was different in various malignant tumors. TSR can be useful for predicting prognosis through a routine microscopic examination of malignant tumors.


Introduction
Pathological examination is performed through the interpretation of glass slides with hematoxylin and eosin (H&E) staining. In the pathological assessment of malignant tumors, the primary tumor, regional lymph node, and distant metastasis are evaluated according to the American Joint Committee on Cancer (AJCC) Cancer Staging Manual [1]. Tumor differentiation, lymphovascular invasion, perineural invasion, and resection margin involvement are evaluated in daily practice. Useful parameters for predicting the patient's prognosis should have easy identification, high reproducibility, and less discrepancy between investigators. An epithelial tumor is composed of the tumor and surrounding stroma. Interaction between tumor cells and intra-and peritumoral stroma is important in tumor progression [2]. Evaluating these interactions can be useful for understanding tumor behavior. Stroma includes various components, such as immune cells, fibroblasts, and the extracellular matrix [3][4][5][6]. The tumor-stroma ratio (TSR), defined as the proportion of tumor area in the overall tumor, has been studied as a histologic assessment [2,. 2 of 13 TSR is assessed by microscopic observation with H&E staining and is a method that can be sufficiently evaluated in routine pathology laboratories. However, the impact of the proportion of stroma is not clear in terms of whether it accelerates or suppresses tumor progression. Recently, the prognostic implications of TSR have been exhibited for various malignant tumors [2,. In colorectal cancers, low stroma was associated with less frequent vascular and perineural invasion and distant metastasis [45]. HIF-1α was found to be highly expressed in stroma-high tumors, with correspondingly high microvessel density in colorectal cancers. There is no conclusive information on the prognostic impacts of TSR in various malignant tumors. TSR is divided into high TSR (stroma-low) and low TSR (stroma-high) by evaluation criteria. In previous studies, the evaluation criteria of TSR affected high TSR rates [2,. If the evaluation criteria are different, the prognostic implications of TSR can differ. The evaluations of TSR have been shown to have good interobserver agreement [58] but may be more influenced by criteria. In addition, there are carcinomas that require more careful evaluation, such as lung cancer, where the amount of stroma can be inherently different between histological subtypes. It is difficult to evaluate the implications of TSR from individual studies. It is important to determine the direction of further research and analysis from a comprehensive analysis. A meta-analysis study using previous literature can be useful in obtaining comprehensive information.
We investigated high TSR rates of various malignant tumors according to malignant tumor evaluation criteria. The correlations between TSR and survival were elucidated through the subgroup analysis based on malignant tumors. The high TSR rates and prognostic impact of TSR according to evaluation criteria were analyzed.

Published Study Search and Selection Criteria
Relevant articles were obtained by searching the PubMed database through 15 February 2023. The following keywords were used in the search: "(tumor-stroma ratio or carcinoma-stroma ratio) AND (cancer or tumor or malignancy or neoplasm or carcinoma) AND (prognosis or prognostic or survival)". The titles and abstracts of all searched articles were screened for inclusion and exclusion. Included articles had information on the correlation between TSR and survival in malignant tumors. However, non-original articles, such as case reports and review articles, were excluded. Articles not written in English were not included in the present study. Finally, 51 eligible articles were included in the meta-analysis (Table 1). The PRISMA checklist is shown in Supplementary Table S1. In addition, we evaluated eligible studies using the Newcastle-Ottawa Scale, and the results are presented in Table 2.

Data Extraction
All data were extracted from 51 eligible studies [2,. Extracted data included the author's information, study location, number of patients analyzed, and high TSR evaluation criteria. The number and survival rates of high and low TSR were also investigated. For the quantitative aggregation of the survival results, the correlation between TSR and survival was analyzed according to the hazard ratio (HR) using one of three methods. In studies that did not take note of HRs or confidence intervals (CIs), these variables were calculated from the presented data using the HR point estimate, the log-rank statistic or its p-value, and the O-E statistic (the difference between the number of observed and expected events) or its variance. If those data were unavailable, HR was estimated using the total number of events, the number of patients at risk in each group, and the log-rank statistic or its p-value. Finally, if the only useful data were in the form of graphical representations of survival distributions, survival rates were extracted at specified times to reconstruct the HR estimate and its variance under the assumption that patients were censored at a constant rate during the time intervals [59]. The published survival curves were read independently by two authors to reduce reading variability. The HRs were then combined into an overall HR using Peto's method [60]. Two independent authors obtained all data (Pyo J.S. and Kim N.Y.).

Statistical Analyses
The meta-analysis was performed using the Comprehensive Meta-Analysis software package (Biostat, Englewood, NJ, USA). The high TSR rate was investigated in various malignant tumors. TSR's prognostic impact was evaluated, dividing survival into overall survival (OS) and disease-free survival (DFS). Heterogeneity between the studies was checked by the Q and I 2 statistics and expressed as p-values. Additionally, sensitivity analysis was conducted to assess the heterogeneity of eligible studies and each study's impact on the combined effects. In the meta-analysis, because the eligible studies used various malignant tumors and populations, a random-effects model rather than a fixedeffects model was more suitable. Begg's funnel plot and Egger's test were used; if significant publication bias was found, the fail-safe N and trim-fill tests were also used to confirm the degree of publication bias. The results were considered statistically significant at p < 0.05.

•
A primary search using the PubMed database found 509 relevant articles. In screening and reviewing, 409 were excluded due to inapplicable or insufficient information. Among the remaining articles, 49 reports were excluded for the following reasons: non-original articles (n = 31), non-human studies (n = 5), a language other than English (n = 11), and articles including duplicated patients (n = 2) (Figure 1).

Selection and Characteristics of the Studies
• A primary search using the PubMed database found 509 relevant articles. In screening and reviewing, 409 were excluded due to inapplicable or insufficient information. Among the remaining articles, 49 reports were excluded for the following reasons: non-original articles (n = 31), non-human studies (n = 5), a language other than English (n = 11), and articles including duplicated patients (n = 2) (Figure 1).

Correlation between High Tumor-Stroma Ratio and Survival
In overall cases, high TSR was significantly correlated with better OS and DFS compared to low TSR (HR 0.631, 95% CI 0.542-0.734, and HR 0.564, 95% CI 0.476-0.669, respectively; Tables 4 and 5). Significant correlations with OS were found in breast, cervical, colorectal, esophageal, head and neck, ovary, stomach, and urinary tract cancers. In addition, there were significant correlations of DFS in breast, cervical, colorectal, esophageal, laryngeal, lung, and stomach cancers. However, in endometrial and pancreas cancers, high TSR was significantly correlated with a worse prognosis. In subgroup analysis based on evaluation criteria, there were significant correlations between high TSR and better OS and DFS in the subgroups with criteria <50% and 50%. In the subgroup with criteria >50%, patients with high TSR had a better OS, but not DFS, compared to patients with low TSR.

Discussion
In the present meta-analysis, the rates of high TSR were evaluated in various malignant tumors. In addition, the correlation between TSR and survival was investigated through a meta-analysis. Previous studies used variable methods for evaluating TSR. The criterion for high TSR is usually 50% through visual inspection. Therefore, a meta-analysis is more useful for understanding the prognostic implication of TSR. The present study is the first meta-analysis, to the best of our knowledge, to elucidate the prognostic impacts of TSR according to malignant tumors and evaluation criteria.
Regardless of the origin of epithelial tumors, malignant tumors initiate through invasion into the basement membrane and progress to the stroma. This process induces changes in the characteristics of the stroma, including fibroblast proliferation and extracellular matrix deposition, through the production of cytokines and enzymes with the surrounding stroma [61][62][63][64]. Therefore, in malignant tumors, the interaction between tumor cells and stroma is important [33]. Malignant tumors have intratumoral stroma and an interface with peritumoral stroma. The definition of TSR is the proportion of tumor area in the overall tumor, including the stroma. Tumors with low TSR, which have abundant stroma, are considered active interactions between tumor cells and stroma. The tumor growth and progression are associated with the tumor microenvironment [65]. However, the detailed evaluation of the tumor environment through hematoxylin and eosin staining can be limited. TSR can be considered as a simplified analysis of the interaction between tumor cells and stroma. Therefore, the assessment of TSR may be applicable for predicting the prognosis through the routine evaluation of histology.
Recently, assessments using image analyzers have been increasingly used in research and practice. Evaluation of TSR is performed through various methods, including eyeballing and the use of a digital image analyzer [15]. The assessment of TSR can be affected by multiple factors, including the discrepancy between investigators. The evaluation criteria for high TSR are yet to be elucidated. To diminish the discrepancy caused by various factors, an image analyzer is used for evaluating TSR. The value of TSR can be different according to the evaluation foci within the tumor. The evaluation area can also affect the value of TSR, and two-tier or three-tier classification can affect the prognostic impact of TSR. Further cumulative studies for the prognostic implication of TSR gradients by evaluation criteria will be needed.
Most eligible studies investigated the evaluation criterion of 50% for high TSR. However, the previous meta-analysis showed no results for evaluation criteria [66]. With the increasing cut-off for high TSR, the rate of high TSR is lowering. The rates of high TSR were 0.624, 0.609, and 0.399 in the <50%, 50%, and >50% cut-off subgroups, respectively. In the present study, patients with high TSR had better OS and DFS than those with low TSR in the <50% and 50% cut-off subgroups. However, in the subgroup with criteria >50%, patients with high TSR had a better OS, but not DFS. In the assessment of OS in criteria >50%, colorectal cancers are only included. However, in the assessment of DFS in criteria >50%, one breast cancer study and one colorectal cancer study were included. Among these studies, there was no significant correlation between high TSR and better prognosis in a study on only breast cancer [50]. In subgroup analysis, breast cancers showed a significant correlation between high TSR and better DFS (Table 5). Although there may be a difference in the degree of HR, it can be considered that there is no significant difference in the relationship with prognosis according to the criteria.
In a pathological examination, the evaluation criteria of TNM staging differ according to malignant tumors. For example, in lung cancers, the pT stage is evaluated by tumor size and invasion depth [1]. The invasive size, rather than the overall tumor size, was significantly correlated with lung adenocarcinoma [67,68]. In addition, lung adenocarcinoma includes various histologic subtypes, such as lepidic, acinar, micropapillary, papillary, and solid adenocarcinomas. Although these subtypes have variable amounts of stroma, the specific correlation between histologic subtypes and stroma amount is not clear. Evaluating the TSR of lepidic adenocarcinoma, which has similarities with the lung's normal parenchyma, may not be easy. Ichikawa et al. reported that lung adenocarcinoma with low TSR was significantly correlated with favorable tumor behaviors [19]. However, Xi et al. reported a significant correlation between low TSR and worse prognosis [33]. Xi's report included adenocarcinoma and squamous cell carcinoma. However, the prognostic impact based on histologic subtypes of non-small cell lung cancers could not be elucidated in that study. In a previous study, TSR was not correlated with various clinicopathological characteristics, including histologic subtypes, pT stage, pN stage, and pTNM stage [33].
A previous meta-analysis was reported for the prognostic roles of TSR in gastrointestinal tract cancers [66]. There were significant correlations between high TSR and better OS in colorectal, stomach, and liver cancers. However, some discrepancies are present compared to our results. In the current meta-analysis, there was no significant correlation between TSR and OS in liver cancer. The highest and lowest rates of high TSR were found in endometrial cancer (86.5%) and stomach cancer (44.8%), respectively. This discrepancy can be caused by different characteristics of malignant tumors. Endometrial intraepithelial neoplasm, which is a precursor of endometrial cancer, has less stroma compared to the tumor area. Interestingly, for endometrial carcinoma, it was shown that high TSR was significantly correlated with worse OS and DFS. However, cervical cancers showed a significant correlation between TSR and better OS and DFS. In addition, high TSR of ovarian cancers was significantly correlated with better OS. These results suggest that the biology of tumor-stroma interactions may differ amongst cancer types.
There were some limitations in the current meta-analysis. First, high TSR rates based on histologic subtypes of each tumor could not be investigated due to insufficient information. Second, each study was not described for the evaluation area or section. In addition, it is uncertain whether the evaluation foci for TSR are hot spots or representative regions. Third, a comparison between eyeballing and image analyzers could not be performed due to insufficient information on eligible studies. Fourth, we were unable to conduct analyses by criteria subgroup for high TSR for each cancer type due to insufficient information.

Conclusions
In conclusion, our results showed that high TSR rates were different between malignant tumors. High TSR was significantly correlated with better survival rates, although some malignant tumors had no correlation or opposite correlation. Our results show that endometrial and pancreatic cancers are correlated with a poor prognosis. TSR can be useful for predicting prognosis through a routine microscopic examination of malignant tumors.
Further studies for standardized histopathologic criteria will be needed in the application of TSR.