Microsatellite instability leads to poor prognosis in patients with early-stage endometrial cancer? a meta-analysis

Background Poor prognosis of early-stage endometrial cancer (EC) is often accompanied by microsatellite instability (MSI). We hypothesized that MSI is an independent marker for poor prognosis of early-stage EC. To demonstrate this hypothesis, we evaluated the correlation between MSI and early-stage EC prognosis by meta-analysis. respectively. The disease-free survival (DFS), the overall survival (OS), and the progression-free survival (PFS) were pooled to analyze the correlation between MSI and prognosis in patients with early-stage EC. Besides, Egger's regression and Begg's test were used to detect Publication bias. There were 7 studies met the inclusion criteria and were enrolled in our meta-analysis with a sample size of 1150, and the included patients with early-stage EC were all endometrioid endometrial cancer (EEC). The pooled hazard ratios (HRs) in early-stage EC shows that MSI is signicantly associated with lower DFS [HR = 3.90, 95%CI (2.81–6.99), p = 0.000], OS [HR = 1.48, 95%CI (1.12–1.96), p = 0.006], and PFS [HR = 2.41, 95%CI (1.05–5.52), p = 0.038]. There was no signicant heterogeneity in the studies pooled analysis of DFS, OS, and PFS. There was also no statistical publication bias, the P-value of Egger`s test of OS and DFS is p = 0.535 and p = 0.639 respectively.


Conclusion
MSI is most likely an independent marker of poor prognosis in early-stage EC, and this correlation is even more signi cant in patients with EEC.

Background
Endometrial cancer (EC) is one of the most common cancers in the female reproductive tract, and the increase of incidence and mortality has an up-trend year by year [1,2]. Assessment of prognosis is of great signi cance in the clinical management of EC, and prognostic assessment is key to identifying prognostic markers. [3,4].
Mismatch repair (MMR) contains four proteins: MLH1, MSH2, MSH6 and PMS2. When one or more of these proteins are not expressed, it is called mismatch repair de ciency (MMRd). Due to MMRd, errors produced by DNA replication cannot be repaired in time, known as microsatellite instability (MSI) [5]. MSI is the most sensitive and speci c marker of MMRd, and MMRd can be inferred by examining MSI [6,7].
MSI accounts for 20-40% of patients with sporadic EC and has been associated with endometrioid histology [8,9]. MSI is considered to be an important prognostic marker in the EC. Therefore, an increasing number of studies have focused on the correlation between MSI and prognosis of EC.
Some studies have shown MSI to be associated with better prognosis in EC [10,11], some studies have shown the opposite [12][13][14], and others have shown no correlation with prognosis, including a meta-analysis [15][16][17]. The same contradiction also exists in studies with early-stage EC [18,19]. Since patients with early-stage EC usually do not require adjuvant therapy after surgery, MSI is more strongly correlated with prognosis. However, none of the relevant metaanalyses have been reported.
To clarify the correlation between MSI and prognosis of EC, we performed a meta-analysis that included the disease-free survival (DFS), the overall survival (OS), and the progression-free survival (PFS) of early-stage EC.

Data sources and search strategy
This meta-analysis was rigorously evaluated by the Preferred Reporting Items for Systemic Reviews and Meta-Analyses (PRISMA) guideline [20]. Database as PubMed, EMBASE, and the Cochrane Collaboration Library was searched from inception to October 2020, the language restriction was English.
We adjusted the MeSH terms combined with related text words to comply with the relevant rules for searching our interesting studies in each database. Our search strategy was that: (Endometrial Neoplasm or Endometrial Neoplasms or Endometrial Carcinoma or Endometrial Carcinomas or Endometrial Cancer or Endometrial Cancers or Endometrium Cancer or Cancer of the Endometrium or Carcinoma of Endometrium or Endometrium Carcinoma or Endometrium Carcinomas or Cancer of Endometrium or Endometrium Cancers) AND (Mismatch repair or Microsatellite Instability or Replication Error Phenotype or Replication Error Phenotypes) AND survival.

Study selection
Two independent researchers (Jing-ping Xiao and Yun-zi Wang) ltered all the titles and abstracts of the retrieved studies to identify potential studies. The retrieved studies that met the inclusion criteria were evaluated in full text. Each of these discrepancies was resolved through discussion, and if con icts remained, a third reviewer (Ji-sheng Wang) was involved.

Inclusion criteria
Studies containing the correlation between MSI or MMRd and prognosis of EC were included if they met the following criteria. (1) The stage of EC was early (stage I-II); (2) Reported DFS, OS, or PFS associated with MSI or MMRd; (3) Studies directly reported hazard ratios (HRs) with 95% con dence intervals (CIs) or have a Kaplan-Meier survival curves that can be used to extract HRs.
Editorials, meeting reports and letters to the editors were all excluded.

Data extraction
Two researchers independently screened articles following inclusion criteria, and any differences were resolved by consensus. From each study, we extracted study characteristics, baseline characteristics, and pre-established outcomes for DFS, OS, and PFS.

Quality Assessment
Two researchers (Jing-ping Xiao and Yun-zi Wang) separately applied the Newcastle-Ottawa Statement to evaluate the quality of eligible studies, including selection, comparability, and exposure. Nine points were included in the scale, and a score greater than or equal to 7 was considered to be a high-quality study.
A score of 4-6 was considered a good quality study and a score of 3 or less was considered a low-quality study [21], and discrepancies were resolved through discussion, with the involvement of a third reviewer (Ji-sheng Wang) if a con ict remained.

Data synthesis and analysis
A Stata (version 14) software was used to analyze all results. The hazard ratios (HRs) would be extracted and calculated by the Kaplan-Meier survival curves if there was not a directly available HRs in the study. If an I 2 greater than or equal to 50% indicated signi cant heterogeneity, the HRs were merged with the corresponding 95% CIs using a random-effects model; otherwise, the xed-effects model was used. Publication bias was statistically assessed by Egger's regression and Begg's test, where a p-value < 0.05 was considered to be a signi cant publication bias.

Results
Literature search Figure 1 illustrates the ow of the selection of eligible articles. A total of 720 articles were identi ed by searching PubMed, Cochrane, and EMBASE. 469 articles remained after removing duplicate les. Following the scanning of titles and abstracts, 50 articles were selected for full-text reading. Finally, we included seven studies [18,19,[22][23][24][25][26] that met the inclusion criteria for our meta-analysis. Table 1 shows the characteristics of the seven studies included. Of these studies, four studies were conducted in Europe (Spain, Italy, and Norway) [18,19,22,23], one study was conducted in Asia (Korea) [25], two studies was conducted in the Americas (Canada) [24,26]. Five studies directly reported HRs for DFS, OS, or PFS, while HRs of the other two studies [18,25] were extracted from Kaplan-Meier survival curves. Seven studies were cohort studies and one study was a clinical trial. Four studies assessed MSI by using ve recommended quasimonomorphicmononucleotide markers, and three studies assessed MSI by immunohistochemistry testing. As shown in Table 2, all studies scored 7 or higher and were high-quality studies.

Correlation between MSI and DFS in early-stage EC
The pooled HRs in early-stage EC shows that MSI is signi cantly associated with lower DFS [HR=3.90, 95%CI (2.81-6.99), p=0.000], as shown in Figure 2a.

Correlation between MSI and OS in early-stage EC
The pooled HRs in early-stage EC shows that MSI is signi cantly associated with lower OS [HR=1.48, 95%CI (1.12-1.96), p=0.006], as shown in Figure 2b.

Correlation between MSI and PFS in early-stage EC
As shown in Figure 2c, the pooled HRs in early-stage EC shows that MSI is signi cantly associated with lower PFS [HR=2.41, 95%CI (1.05-5.52), p=0.038].

Publication bias
No signi cant publication bias was detected by the funnel plot test (Figure 3). Additionally, there was also no statistical publication bias, the p-values of Egger`s test for DFS and OS are p=0.639 and p=0.535 respectively.

Sensitivity analysis
To explore the sensitivity of the pooled HRs of DFS, OS, and PFS in early EC, we omitted each study individually from the pooled analysis. The exclusion of any study had no signi cant in uence on the results (Figure 4).

Discussion
The correlation between MSI and EC prognosis has been one of the hot topics of studies for more than two decades. Unfortunately, most of the current studies on the correlation between MSI and prognosis of EC have shown inconsistent results. For example, the meta-analysis by Diaz-Padilla et al. showed no correlation between MSI and prognosis in patients with EC [15], Nagle et al. reported that MSI was signi cantly associated with poor prognosis [27], while Black et al. showed that MSI was signi cantly associated with a good prognosis [10]. Therefore, the clinical prognostic signi cance of MSI in EC remains unclear.
By this meta-analysis, we found that a signi cant association between MSI and poorer prognosis in early-stage EC. We analyzed the correlation between MSI and prognosis of early-stage EC by DFS, OS, and PFS, then found that the DFS, OS, and PFS of early-stage EC patients with MSS (microsatellite stability) were signi cantly higher than patients with MSI, which is consistent with the cancer-speci c survival of early-stage EC reported in the study by Bilbao et al. [23].
Also, none of the three pooled forest plots were heterogeneous, so we used a xed-effects model, which proved the high reliability of the results we obtained.
Furthermore, in the study by Black et al. [10], MSI was associated with a good prognosis for EC, which is the opposite of our ndings. The reason for the analysis maybe that 20% of the patients included in Black's study were non-endometrioid endometrial cancer (EEC). In contrast, in our meta-analysis, the included patients with early-stage EC were all EEC, which may indicate that MSI can be signi cantly associated with worse prognosis only in EEC patients, which is also consistent with the study by Nagle et al. [27].
Usually, the majority of patients with EC in all clinical stages (stages I-IV) received one or more adjuvant therapies, which increase the uncertainty as to whether MSI has a prognostic predictive role. Whereas, in our study, women with early-stage EC generally did not receive adjuvant therapy after surgery. Thus, our study is better able to exclude the confounding effects of adjuvant therapy and illustrate the correlation between MSI and EC.
There are still some de ciencies in our study. First, some of the data came from the extraction of survival curves, which may produce some deviations compared with the real data. Second, the number of studies included in the pooled PFS was small, and more studies are needed to support our conclusions. Third, the vast majority of the studies we included were retrospective case studies, which carries the risk of selective reporting. Fourth, the four studies used genotyping for MSI detection, and other three of the included studies used immunohistochemistry for MSI detection, but so far, the concordance between the two detection methods has not been ascertained in EC.

Conclusions
The results of our meta-analysis showed that MSI is most likely an independent marker of poor prognosis in early-stage EC,and this correlation is even more signi cant in patients with EEC. This correlation requires more large-scale, well-designed prospective studies as well as randomized controlled trials to illustrate the mechanisms of the relationship between MSI and early EC.  * If there is a positive symbol that means score one point; ** A negative symbol means no point.