Diagnostic and Prognostic Roles of Thrombospondin-2 in Digestive System Cancers

Background Cancers of digestive system have high case-fatality rate. It is important to find more appropriate methods in diagnosing and predicting gastrointestinal malignances. And thrombospondin-2 (TSP-2) was reported to have the functions, although results were not identical. So we performed this meta-analysis to clarify the significance of TSP-2 in this area. Methods PubMed, Embase, Web of Science, Cochrane Library, and Clinicaltrial.gov were searched for relevant studies. Data were extracted from these involved records. For the meta-analysis of diagnostic test, bivariate mixed effect model was used to estimate diagnostic accuracy. For prognosis part, HRs and their 95% CIs were pooled to compare the overall survival (OS) and disease-free survival (DFS) between patients with high TSP-2 and low TSP-2. Results Nine records were eligible for the analysis of diagnostic test. Pooled results were as follows: sensitivity 0.60 (0.52, 0.68), specificity 0.96 (0.91, 0.98), positive likelihood ratio (PLR) 15.4 (7.3, 32.2), negative likelihood ratio (NLR) 0.42 (0.34, 0.50), and diagnostic odds ratio (DOR) 37 (18, 76). While in prognosis part, 10 articles were included. Patients with increased TSP-2 had shorter OS (HR = 1.64, 95% CI = 1.21-2.22); however, no difference was found in DFS between TSP-2 high and low groups (HR = 1.44, 95% CI = 0.28-7.33). Conclusions TSP-2, as a diagnostic marker, has a high specificity but a moderate sensitivity. Meanwhile, it plays a role in predicting OS. Therefore, making TSP-2 a routine assay could be beneficial to high-risk individuals and patients with digestive malignances.


Introduction
Cancer is a global health problem [1]. According to the cancer statistics in 2022, around 343,040 persons were newly diagnosed with cancers of digestive system, and 171,920 patients died of it in America. Cancers of the pancreas and liver including intrahepatic bile duct have the lowest 5-year survival rates at only 11% and 20%, respectively [2]. And it was predicted that pancreas cancer would become one of the top cancer killers by 2030 [3]. Therefore, finding effective methods assisting in early diagnosis and accurate monitoring is necessary for better survival.
Thrombospondin (TSP), a matricellular glycoprotein, was first discovered by Baenziger et al. in 1971 [4]. TSP-2 is one of the thrombospondin protein families, which functions by mediating cell-matrix and cell-cell interactions [5,6]. It is mainly generated by smooth muscle cells and fibroblasts and has an influence on several mammalian biology, such as cell proliferation and migration, tumor angiogenesis, and wound healing [5,7].
During tumorigenesis, the expression of TSP-2 is changed [8,9]. Researchers believe that TSP-2 has diagnostic and prognostic value, but conclusions remain inconsistent [10,11]. Thus, we carried on a meta-analysis assessing the functions of TSP-2, hoping to provide some suggestions for doctors in diagnosing and predicting the cancers.
In this study, the analysis can be divided into diagnostic and prognostic components. In the first part, we assessed the diagnostic accuracy including sensitivity, specificity, likelihood ratio (LR), and area under the curve (AUC). As for the prognosis part, we compared OS and DFS in TSP-2 high and low groups.

Search Strategy.
PubMed, Embase, Web of Science, Cochrane Library, and Clinicaltrial.gov were searched for relevant studies up to January 20, 2022. The search term included "(Thrombospondin-2) or (TSP-2) or (THBS2)" in Embase, Web of Science, Cochrane Library, and http://Clinicaltrial .gov and MeSH term relating to "Thrombospondin-2" in PubMed. Restricted to the inclusion and exclusion criteria, two reviewers (F.G. and W.C.) screened the records, and discrepancies were reconfirmed by a third reviewer (H.C.).

Selection Criteria.
Inclusion criteria for diagnosis part are as follows: (1) patients diagnosed with digestive system cancers, (2) studies assessing diagnostic accuracy of blood TSP-2, and (3) studies with negative (FN). Studies met the following criteria would be included in the prognosis part: (1) clinical studies (both prospective and retrospective), (2) patients included diagnosed with digestive system cancers, (3) studies with comparisons of TSP-2 regarding DFS or OS, and (4) results presented in the form of HR and 95% CI. Exclusions for both parts were as follows: (1) cell studies or animal experiments, (2) replicate studies or studies without full text, and (3) studies lack of interested results or data.
2.3. Data Extraction. Data pertaining to author, journal, location, disease, sample size, and cut-off value were extracted. TP, FP, TN, and FN were collected from studies of diagnostic test; HRs and their 95% CIs were from prognosis studies. We used Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) and Newcastle-Ottawa Scale (NOS) to assess the quality of the records relating to diagnosis and prognosis, respectively.

Statistical Analysis.
We used STATA version 16.0 (Stata-Corp, College Station, TX, USA) for analysis. In the metaanalysis of diagnostic test, specificity, sensitivity, LR, DOR, and AUC were pooled by bivariate mixed effect model. Fagan plots were used to calculate the posttest probability, and Deek's funnel plot was conducted to check the publication bias.
As for the prognosis part, we pooled HRs and their 95% CIs to compare OS and DFS between high TSP-2 and low TSP-2 groups. To explain the relationship of TSP-2 and clinicopathological feature of the patients, OR was calculated. Begg's test and Egger's test were performed to evaluate the publication bias. I 2 statistic and chi-squared test were used to assess the heterogeneity. When I 2 ≤ 50% and P > 0:1, the heterogeneity was considered acceptable. Otherwise, random-effects models were employed, and sensitivity and subgroup analyses were performed to account for apparent heterogeneity.

Results
3.1. Study Selection. After searching 5 databases mentioned above, a total of 1611 records were identified (Figure 1), of which 503 were duplicates, leaving 1108 for initial screen. 51 were left for further evaluation after screening titles and abstracts. Following detailed reading, 16 relevant studies were included in this meta-analysis. Of these records, 9 are related to diagnosis, and 10 are related to prognosis (3 articles contained both data).

Diagnosis Part
3.2.1. Study Characteristics and Quality Assessment. Main characteristics of 9 included articles were presented in Table 1. A total of 2231 subjects were involved in this meta-analysis of diagnostic test, including 1307 patients diagnosed with digestive cancers and 924 controls. TSP-2 was detected in their blood by enzyme-linked immunosorbent assay (ELISA). Cut-off values varied from 14.85 ng/ml to 42 ng/ml, and two studies did not tell their cut-offs. Results of the quality evaluation according to QUADAS 2 were showed in Supplementary Fig. S1. The risk of bias and the application concerns were at an acceptable level. The Deek's plot was fairly symmetrical, indicating that there was no publication bias either (P = 0:53, Supplementary  Fig. S2).

Prognosis Part
3.3.1. Study Characteristics and Quality Assessment. Table 2 depicts the characteristics of 10 prognosis studies. There were 1526 individuals involved totally. Six studies tested blood TSP-2, while others detected TSP-2 in tissue. Nine records provided HRs and their 95% CIs regarding OS, while only 3 have the data of DFS. The NOS score of adopted articles was greater than 4 (the total score is 9 points, Supplementary Table S1).

Disease-Free
Survival. Data were available in 3 articles, and 275 subjects were analyzed. By pooling the provided HRs of DFS, no significant difference was found between the high TSP-2 and low TSP-2 groups (HR = 1:44, 95% CI = 0: 28-7.33). And the forest plot was shown in Supplementary Fig. S4. Further sensitivity analysis and subgroup analysis were not carried on because of the limited number of eligible studies.

Relationship of TSP-2 and Clinicopathological
Characteristics. Table 5 describes the associations between

Discussion
As an essential matricellular protein, TSP-2 plays a complicated role in some disease including inflammation, fibrosis, and malignances [12,13]. It is regarded as a protector in     [14]. A study carried on TSP-2 deficient mice, suggesting that TSP-2 could limit inflammatory cell infiltration during delayed-type hypersen-sitivity [15]. Moreover, TSP-2 can accelerate fibrosis and is considered as a biomarker for liver fibrosis [16][17][18][19]. Interestingly, although TSP-2 could hinder angiogenesis, whether it is beneficial or not during tumorigenesis is controversial. Some researchers discovered that TSP-2 could inhibit     [20]. Conversely, some studies clarified that TSP-2 promoted cancer growth and invasion in pancreatic cancer [21,22]. And in intrahepatic cholangiocarcinoma, researchers found elevated levels of TSP-2, inhibiting angiogenesis and promoting lymphangiogenesis, leading to rapid cancer spread [23].
Researchers have already focused on the diagnostic role of TSP-2 in digestive cancers [24]. Kim et al. regarded TSP-2 as an appropriate predictor for early pancreatic ductal adenocarcinoma. TSP-2 combined with CA19-9 could achieve specificity at 98% and sensitivity at 87% [25]. And in the gastric cancer, colorectal cancer, and hepatocellular carcinoma, TSP-2 also aided in diagnosis, especially in combination with other markers [10,26,27]. On the other hand, it was reported that TSP-2 was related to survival rates of patients with digestive system cancer, but the results were not identical. In colorectal cancer, some researchers found that patient with high TSP-2 had inferior OS [11]. Some researchers found that the difference in survival between the high and low TSP-2 groups was not statistically significant [10], while others found those with elevated TSP-2 had a longer DFS [28]. And in gastric cancer, OS rates were higher in the high TSP-2 group [29]. The reason may be that TSP-2 can promote the progression of Helicobacter pyloriassociated gastric cancer [30].
To our knowledge, this is the first meta-analysis assessing the significance of TSP-2 in diagnosing or predicting gastrointestinal cancers. We found that TSP-2 achieved high overall specificity at 0.96 (0.91, 0.98); however, the sensitivity was moderate. The AUC of 0.83 (0.80, 0.86) for SROC was also indicated a satisfactory diagnosis. Most of the involved papers focused on pancreatic cancer, indicating the importance of TSP-2 in the diagnosis of pancreatic cancer, as was reported in Cancer Discovery [31]. And all of them detect TSP-2 in blood by ELISA. Samples are easy to get without invasion, and the cost is low. Therefore, we regard TSP-2 as a good marker for early screening, especially for those at a high risk of pancreas cancer.
As for the role of TSP-2 in prognosis, the pooled HR indicated that patients with overexpressed TSP-2 had shorter OS. But there was no difference in the DFS rate in the two targeted groups. Subgroup analysis illustrated increased levels of TSP-2 in tissue and blood, and both samples had prognostic value. The relationship with clinicopathological characteristics revealed that TSP-2 increased when tumor staging progressed, which is in consistent with the conclusion TSP-2 increased as cancer developed [21].
How TSP-2 works in gastrointestinal cancers remains unknown. Some researchers believed that inhibition of TSP-2 promoted epithelial mesenchymal transition in gastric cancer, resulting in the progression. But the detailed mechanism was not explained [32]. A study elucidated that TGF-β1 could promote the expression of TSP-2 which in turn interacted with integrin αvβ3/CD36, and then activated MAPK signaling to drive the progression of pancreatic cancer [21]. And in colorectal cancer, Xu et al. put forward that TSP-2 accelerated cancer cell proliferation by adjusting the Toll-like receptor 4 (TLR4) [33]. However, there is still a long way to go to unearth the mechanisms of TSP-2 in cancers.
Our study has some limitations. Different samples and cut-offs increased heterogeneity. Some conference abstracts and articles without available full text were not included, leading to publication bias to a certain extent. The number of eligible records involved in our study was not very large. Therefore, articles adopted in each subgroup were limited when subgroup analysis was performed.

Conclusions
We attach great importance to TSP-2 in the diagnoses and predictions of digestive malignances. TSP-2 alone or in combination with other markers has satisfactory diagnostic accuracy and prognostic value. And TSP-2 in the blood is easily detected, noninvasive, and inexpensive. Therefore, it is strongly recommended to add TSP-2 to the routine detection of tumor markers.

Data Availability
All data generated or analyzed during this study are included in this published article.