Prognostic role of STAT3 in solid tumors: a systematic review and meta-analysis

1Signal transducer and activator of transcription proteins 3 (STAT3) is well demonstrated to play a crucial role in tumor development and cancer-related inflammation. STAT3 is also linked to inflammation-related oncogenesis initiated by genetic alterations and environmental factors, and is constitutively activated in various cancers. Persistent activation of STAT3 is involved in promoting tumor cell proliferation, survival, tumor invasion, angiogenesis and immunosuppression, inducing and maintaining a pro-carcinogenic inflammatory microenvironment. Growing studies identified novel tumor-promoting functions of STAT3 in mitochondria metabolism, drug resistance, epigenetic regulation, cancer stem cells and pre-metastatic niches. Given the pivotal role in tumor development, STAT3 represents an attractive therapeutic target for solid tumors. Recently, accumulating studies have demonstrated STAT3-targeted therapy could effectively restrain tumor development in various solid tumors. However, the prognostic value of STAT3 overexpression in human solid tumors is still controversial.

2A plenty of studies showed that elevated STAT3 expression in tumor tissue was correlated with poor survival of patients with various solid tumors such as gastric cancer, lung cancer, gliomas, colorectal cancer, ovarian cancer, cervical cancer, hepatocellular carcinoma, melanoma, esophageal cancer, osteosarcoma, pancreatic cancer, thymic epithelial tumor, astrocytomas, lingual squamous cell carcinoma, nasopharyngeal carcinoma, prostate cancer, renal cell carcinoma and Wilms' tumor. However, other studies reported that overexpression of STAT3 was correlated with favorable outcome of patients with breast cancer, gastric cancer, lung cancer, colorectal cancer and melanoma.

3Therefore, we carried out a meta-analysis combining available evidences to evaluate the prognostic value of STAT3 expression in solid tumors. We also evaluated whether the clinical outcome of patients with solid tumors differed between STAT3 phosphorylation state and between different tumor types. This meta-analysis intended to assess the role of STAT3 in relation to survival in solid tumors, thereby supporting more rational development of therapeutic strategies against STAT3.

4Sixty-three studies with a total of 9449 patients were included (Figure 1). Characteristics of included studies are shown in Table 1. Eleven studies evaluated lung cancer, nine evaluated gastric cancer, five evaluated breast cancer, five evaluated gliomas, four evaluated colorectal cancer, three evaluated ovarian cancer, three evaluated cervical cancer, two evaluated hepatocellular carcinoma, two evaluated melanoma, two evaluated esophageal cancer, two evaluated osteosarcoma, two evaluated pancreatic cancer, two evaluated thymic epithelial tumours, two oral cancer, and one each evaluated astrocytomas, chordoma, head and neck squamous cell carcinoma, lingual squamous cell carcinoma, nasopharyngeal carcinoma, pharyngeal cancer, prostate cancer, renal cell carcinoma, and Wilms' tumor. Of these 63 studies, 20 studies evaluated STAT3, 37 studies evaluated p-STAT3, and 6 studies evaluated both STAT3 and p-STAT3. As for the region, 39 studies were conducted in Asia, 13 studies in America, 10 studies in Europe, and 1 study in Austria.

5Antibodies, detection and definition method, and cut-off values of STAT3 expression used in the included studies is summarized in Table 2. Diverse antibodies were used for the assessment of STAT3 expression by IHC. For anti-STAT3 antibody, three studies used clone sc-8019, one study each used clone RB-9237, F-2, sc-7179, 79D7, 124H6, and sixteen studies did not report the antibody clone. For anti-p-STAT3 antibody, eight studies used clone D3A7, four studies used clone sc-7993, two studies used clone 9131, one study each used sc-483, sc-8001, sc-8059, ZP-0647, and twenty studies did not report the antibody clone. The median expression of STAT3 in solid tumors was 47.79%, range from 19.65% to 94.66%.

6The combined analysis of 54 studies showed that STAT3 overexpression in tumor tissue was associated with worse 3-year OS of solid tumors (OR = 2.06, 95% CI = 1.57 to 2.71, P < 0.00001) (Figure 2). There was significant heterogeneity among studies (Cochran's Q P < 0.00001, I2 = 81%), so we conducted meta-regression analysis and subgroup meta-analysis to investigate the possible source of the heterogeneity among studies.

7In the stratified analysis by tumor types, STAT3 expression was associated with worse 3-year OS of gastric cancer (OR = 4.06, 95% CI = 1.86 to 8.89, P = 0.0004), lung cancer (OR = 2.22, 95% CI = 1.31 to 3.77, P = 0.003), gliomas (OR = 4.10, 95% CI = 1.50 to 11.23, P = 0.006), hepatic cancer (OR = 3.75, 95% CI = 1.71 to 8.21, P = 0.001), osteosarcoma (OR = 3.94, 95% CI = 1.83 to 8.51, P = 0.0005) and prostate cancer (OR = 11.08, 95% CI = 1.24 to 98.96, P = 0.03) (Figure 3). There was no significant association between STAT3 expression and 3-year OS of colorectal cancer, ovarian cancer, pancreatic cancer, cervical cancer, melanoma and thymic epithelial tumor (Supplementary Figure S1). Interestingly, STAT3 overexpression was associated with favorable 3-year OS of breast cancer (OR = 0.51, 95% CI = 0.35 to 0.74, P = 0.0004) (Supplementary Figure S2).

8Meta-regression analysis showed that publication year, country, gender and NOS score did not contribute to the heterogeneity (data not shown).

9Analysis of 49 studies showed STAT3 expression was also associated with worse 5-year OS (OR = 2.00, 95% CI = 1.53 to 2.63, P < 0.00001) (Figure 4) of solid tumors. There was also high heterogeneity among studies for 5-year OS (Cochran's Q P < 0.00001, I2 = 82%), so we conducted subgroup meta-analysis.

10Subgroup analysis showed that STAT3 expression was associated with worse 5-year OS of gastric cancer (OR = 5.48, 95% CI = 2.14 to 14.01, P = 0.0004), hepatic cancer (OR = 2.48, 95% CI = 1.41 to 4.35, P = 0.002), osteosarcoma (OR = 4.84, 95% CI = 2.23 to 10.50, P < 0.0001), pancreatic cancer (OR = 9.71, 95% CI = 1.80 to 52.41, P = 0.008) and prostate cancer (OR = 8.35, 95% CI = 1.81 to38.51, P = 0.007) (Figure 5). There was no significant association between STAT3 expression and the 5-year OS of colorectal cancer, lung cancer, ovarian cancer, cervical cancer, melanoma and thymic epithelial tumor (Supplementary Figure S3). STAT3 overexpression was associated with favorable 5-year OS of breast cancer (OR = 0.57, 95% CI = 0.37 to 0.89, P = 0.01) (Supplementary Figure S4).

11Twenty studies evaluated STAT3, 38 studies evaluated p-STAT3 and 5 studies evaluated both STAT3 and p-STAT3. Our result showed that both STAT3 and p-STAT3 overexpression were associated with worse OS of solid tumors. However, elevated p-STAT3 (OR = 2.45, 95% CI = 1.73 to 3.46, P < 0.00001) expression in tumor tissue seemed to be more significantly associated with worse 3-year OS than STAT3 expression (OR = 1.72, 95% CI = 1.10 to 2.70, P = 0.02) (Supplementary Figure S5). Similar result was observed for 5-year OS analysis (Supplementary Figure S6). A subgroup meta-analysis of studies evaluated both STAT3 and p-STAT3 shown that p-STAT3 expression was associated with worse 3-year and 5-year OS of solid tumor, but not STAT3 (Supplementary Figure S7).

12We also evaluated the correlation between STAT3 overexpression and the TNM stage of tumor. High expression level of STAT3 was significantly associated with advanced TNM stage (OR = 0.42, 95% CI = 0.31 to 0.58, P < 0.00001) (Figure 6).

13Next, we conducted subgroup analysis according to STAT3 expression level. Results showed STAT3 expression was associated with poor 3-year OS in the studies using cutoff values of 10%–30% (OR = 3.61, 95% CI = 2.42 to 5.39, P < 0.00001) and 50% (OR = 2.14, 95% CI = 1.29 to 3.57, P = 0.003) (Figure 7) to determine STAT3 positivity. Similar result was observed in 5-year OS (Supplementary Figure S6). However, the studies used cutoff value of STAT3 overexpression as more than 1%-6% tumor cells positive was not associated with 3-year and 5-year OS of solid tumors.

14Meta-analysis of 10 studies showed that STAT3 expression was associated with statistically significant poor 3-year DFS (OR = 3.52, 95% CI = 1.85 to 6.71, P = 0.0001) (Figure 8A) and poor 5-year DFS (OR = 3.37, 95% CI = 1.67 to 6.80, P = 0.0007) (Figure 8B).

15Removal of the studies that was an outlier (score, IRS, > 50% vs 1%–6% for other studies) or no report (NR) with regard to the cutoff of STAT3 overexpression by IHC did not influence results for 3- or 5-year OS (OR = 2.45, 95% CI = 1.73 to 3.48, p < 0.00001; OR = 2.08, 95% CI = 1.46 to 2.96, p < 0.0001; respectively). Exclusion of these studies did not reduce heterogeneity for 3- or 5-year OS (Cochran's Q P < 0.00001, I2 = 82%; Cochran's Q P < 0.00001, I2 = 84%, respectively).

16Removal of studies with NOS score 6 did not influence results for 3- or 5-year OS (OR = 2.49, 95% CI = 1.86 to 3.34, p < 0.00001; OR = 2.33, 95% CI = 1.74 to 3.12, p < 0.00001, respectively). Exclusion of these studies did not reduce heterogeneity for 3- or 5-year OS (Cochran's Q P < 0.00001, I2 = 80%; Cochran's Q P < 0.00001, I2 = 82%, respectively).

17Funnel plot analysis showed that there was no statistical evidence of publication bias in our meta-analysis (data not shown).

18This meta-analysis is the most comprehensive assessment of the literatures regarding STAT3 expression and tumor prognosis to date. We systematically evaluated survival data for 9449 solid tumor patients included in 63 different studies. Our study demonstrated that the expression of STAT3 is a marker of poor prognosis in solid tumors, with consistent results of OS at 3 and 5 years. Regarding to the tumor types, elevated STAT3 expression in tumor tissues were associated with worse OS of gastric cancer, lung cancer, gliomas, hepatic cancer, osteosarcoma, prostate cancer and pancreatic cancer. However, elevated STAT3 expression was associated with better prognosis of breast cancer. In addition, expression level of phosphorylated STAT3 was more significantly associated with worse outcome of solid tumors than unphosphorylated STAT3.

19Our study found there is no significant correlation between STAT3 overexpression and OS of colorectal cancer and ovarian cancer. And STAT3 overexpression in breast cancer tissue is associated with favorable OS. However, recent studies demonstrated that STAT3-targeted inhibitor could restrain tumor development in various solid tumor models including breast cancer, melanoma and ovarian cancer. These divergences suggest that further study is needed to shed more light on the underling mechanism of STAT3 signal pathway in pro-tumor microenvironment in different tumor types.

20There are several important implications in this meta-analysis. First, it shows that STAT3 expression is related to adverse outcome of most solid tumors. Second, it identifies a subgroup of tumors with unfavorable outcome in gastric cancer, lung cancer, hepatic cancer, prostate cancer and glioblastoma, but with favorable outcome in breast cancer. Finally, it emphasizes the potential of STAT3 to developing a valuable therapeutic target and prognostic biomarker for solid tumor.

21This study also has some limitations. First, from the literature we could only extract summarized population-level data rather than individual patient-level data. Second, the method for assessing STAT3 expression and definition of STAT3 positivity are inconsistent. Finally, substantial heterogeneity observed across included studies cannot be fully accounted for by our use of appropriate meta-analytic techniques with random-effects modeling.

22In summary, STAT3 expression in solid tumor tissues is associated with poor survival in most solid tumors, which suggests that STAT3 is a valuable prognostic biomarker and a promising therapeutic target for solid tumors.

23This meta-analysis was conducted according to the statement for reporting systematic reviews and meta-analyses. This study summarized and analyzed the results of previous studies, so the ethical approval was not necessary.

24An electronic search of Pubmed, Web of Science and EBSCO were undertaken for studies evaluating STAT3 or p-STAT3 expression and clinical outcome in solid tumors from 1994 to August 2015. The search was performed with subject heading terms including “signal transducer and activator of transcription 3” or “STAT3 transcription factor” or “STAT3” or “phosphorylated signal transducer and activator of transcription 3” or “phosphorylated STAT3 transcription factor” or “phospho-STAT3” and “neoplasms” and the results were limited to human studies of solid tumors. In addition, the entry “signal transducer and activator of transcription 3” or “STAT3 transcription factor” or “STAT3” or “phosphorylated signal transducer and activator of transcription 3” or “phosphorylated STAT3 transcription factor” or “phospho-STAT3” and the name of each specific solid tumor were used for additional studies. A total of 3547, 3542 and 2914 entries were identified, respectively. Inclusion criteria were the measurement of STAT3 and (or) p-STAT3 by immunohistochemistry (IHC), availability of survival data for at least 3 years, and original article written in English. Exclusion criteria were studies evaluating gene expression of STAT3 measured by polymerase chain reaction (PCR) and STAT3 expression in lymph node and myeloid cells. Citation lists of retrieved articles were manually screened to ensure sensitivity of the search strategy. Study selection was based on the association of STAT3 and survival. Two reviewers (Pin Wu and Dang Wu) evaluated independently all of the full articles for study eligibility. Inter-reviewer agreement was assessed using Cohen's kappa coefficient. Disagreement was resolved by consensus.

25Overall survival (OS) and disease free survival (DFS) were the primary endpoints of interest. Data were extracted using predefined abstraction forms. The following details were extracted by two authors (Pin Wu and Dang Wu): name of first author, year of publication, country of publication, tumor type, patient number, tumor stage, antibodies used for the evaluation, method and score for STAT3 assessment, and cut-off values to determine STAT3 positivity. Data for 3 and 5 year of OS and DFS were extracted from tables or Kaplan–Meier curves for both STAT3 negative and STAT3 positive group.

26The studies included in our meta-analysis were all cohort studies. Two independent authors evaluated the quality of each included study using Newcastle-Ottawa Scale (NOS). The studies with 6 scores or more were considered as high quality studies. A consensus NOS score for each item was achieved finally.

27The relative frequency of OS and DFS at 3 and 5 years between STAT3 negative and STAT3 positive group was presented as an odds ratio (OR) and its 95% confidence interval (CI). Sensitivity analyses were carried out for different analytical methods and cut-offs for defining STAT3 expression and NOS scores for quality assessment of included studies. Publication bias was assessed by visual inspection of the funnel plot.

28Data were extracted from the primary publications and combined into a meta-analysis using RevMan 5.3 analysis software (Cochrane Collaboration, Copenhagen, Denmark). Estimates of ORs were weighted and pooled using the Mantel–Haenszel random effect model. Statistical heterogeneity was assessed using the Cochran's Q and I2 statistics. Differences between subgroups were assessed using methods as previous described by Deeks et al.. Meta-regression analysis was conducted using Stata 12.0 software (StataCorp LP, College Station, TX). All statistical tests were two-sided, and statistical significance was defined as P less than 0.05. No correction was made for multiple statistical testing.