Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Prognostic Significance of Matrix Metalloproteinase-7 in Gastric Cancer Survival: A Meta-Analysis

  • Saeed Soleyman-Jahi,

    Affiliation Cancer Research Center, Cancer Institute of Iran, Tehran University of Medical Sciences, Tehran, Iran

  • Saharnaz Nedjat,

    Affiliation Epidemiology and Biostatistics Department, School of Public Health, Knowledge Utilization Research Centre (KURC), Tehran University of Medical Sciences, Tehran, Iran

  • Afshin Abdirad,

    Affiliation Cancer Research Center, Cancer Institute of Iran, Tehran University of Medical Sciences, Tehran, Iran

  • Niloofar Hoorshad,

    Affiliation Cancer Research Center, Cancer Institute of Iran, Tehran University of Medical Sciences, Tehran, Iran

  • Reza Heidari,

    Affiliation Medical school, Tehran University of Medical Sciences, Tehran, Iran

  • Kazem Zendehdel

    Affiliation Cancer Research Center, Cancer Institute of Iran, Tehran University of Medical Sciences, Tehran, Iran

Prognostic Significance of Matrix Metalloproteinase-7 in Gastric Cancer Survival: A Meta-Analysis

  • Saeed Soleyman-Jahi, 
  • Saharnaz Nedjat, 
  • Afshin Abdirad, 
  • Niloofar Hoorshad, 
  • Reza Heidari, 
  • Kazem Zendehdel


The prognostic role of matrix metalloproteinase-7 in gastric cancer survival has been widely evaluated. However, the results are controversial. We aimed to set up a meta-analysis to reach a conclusion on the prognostic significance of metalloproteinase-7 in gastric cancer survival as well as its association with clinicopathological parameters. We searched popular databases from 1988 until October 2014 to gather eligible peer-reviewed papers addressing the prognostic effect of matrix metalloproteinase-7 in gastric cancer patients' survival. The CASP check list was used for quality appraisal. Pooled hazard ratio (HR) for survival and odds ratio (OR) for association with their 95% confidence interval (CI) were considered as summary measurements. Finally, 1208 gastric cancer patients from nine studies were included in the meta-analysis. Pooled HR estimate for survival was 2.01 (95% CI = 1.62 – 2.50, P < 0.001), which indicated a significant poor prognostic effect for matrix metalloproteinase-7. Sensitivity analysis detected no dominancy for any study. No publication bias was detected according to Egger’s and Begg’s tests. Clinicopathological assessment revealed that higher matrix metalloproteinase-7 expression is associated with deeper invasion (pooled OR = 3.20; 95% CI = 1.14 – 8.96; P = 0.026), higher TNM stage (pooled OR = 3.67; 95% CI = 2.281-5.99; P<0.001), lymph node metastasis (pooled OR = 2.84; 95% CI = 1.89 – 4.25; P<0.001), and distant metastasis (pooled OR = 3.68; 95% CI = 1.85 – 7.29; P<0.001), but not with histological grade. This meta-analysis indicated a significant poor prognostic effect of matrix metalloproteinase-7 in gastric cancer survival. Additionally it was associated with aggressive tumor phenotype.


Despite a recent decline in incidence, gastric cancer (GC) is still the second most frequent cause of cancer-related death worldwide [1]. GC patients are still diagnosed in a late stage and have a poor prognosis. Efficient diagnostic and prognostic modalities seem to be the missing parts in the approach for these patients [2]. Different prognosis observed for patients of the same clinical stage emphasizes the fact that the clinical stage cannot efficiently reflect the biological behavior of the tumor and new biological factors (e.g. biomarkers) are mandatory to complement clinical parameters for more precise decision-making [3].

Matrix metalloproteinase (MMPs) are among the cancer-related biomarkers that have recently attracted notable attention [4]. MMPs are a family of endogenous calcium- and zinc-dependent proteolytic enzymes that are capable of degrading most extracellular matrix (ECM) components, as well as regulating other enzymes, chemokines and even cell receptors. Twenty three types of MMPs have been described so far [5,6].

Many studies have investigated MMPs role in cancer progression. Systematic reviews and meta-analyses of these original reports conclude poor prognostic effects of MMP2 and MMP9 in stomach [7,8], breast [9,10], lung [11,12], colorectal [13,14] and ovarian [15] cancers; additionally they showed clinical significance of MMPs in bladder cancer [16] as well as prognostic effect of MMP7 in colorectal cancer [14]. This body of evidence strongly supports MMPs role in cancer progression.

MMP7, also called Matrilysin, is a distinct family member with proteolytic activity against a wide range of biomolecules including proteoglycans, laminin, fibronectin, casein and more importantly basement membrane collagen type IV [17,18]. It is recognized as pivotal in the MMP family since it activates other MMPs (i.e. MMP-2 and MMP-9) for ECM degradation [19] and possesses the highest activity in the MMP family [20]. Another specific characteristic of matrilysin in contrast to other MMPs is that it is mainly expressed by tumor cells and not by stromal cells [2123]. Other than ECM degradation, MMP7 regulates many other cancer-supporting biochemical processes; it enhances cellular proliferation by increasing insulin-like growth factor and mature heparin-binding epidermal growth factor, cleaves cell to cell contact E-cadherin molecules, inhibits apoptosis in cancer cells [24,25] and induces angiogenesis [26]. Therefore, MMP7 could have a prominent prognostic role in tumors and merits comprehensive investigation. Many studies have assessed MMP7 role in cancer extension.

Elevated levels of MMP7 have been reported in many cancer types (gastric, esophageal, colorectal, pancreatic, prostate, head and neck, lung, hepatocellular and breast), as well as in cancer premalignant lesions (pancreas, stomach, colon, breast and prostate) [27]. In addition, MMP7 has been proposed as a prognostic factor in esophagus squamous cell cancer [23], non-small cell lung cancer [28] and in colorectal [29], breast [30], prostate [31] and urinary and bladder [32] cancers. MMP7 prognostic effect in GC has been widely investigated [3,27,3340]. The original studies about the impact of MMP7 on patients' survival are not consistent [3,27,3339]. A recent meta-analysis demonstrated that MMP7 level is significantly associated with clinicopathological parameters in GC [40]. However, this study did not include survival data such as hazard ratio or risk ratio. Association of biological markers with pathological parameters may or may not be linked to the patient outcome. Therefore, it is important to evaluate whether the observed association of the biomarkers with the baseline variables will affect the patients clinical outcome or not. Thus, we aimed to perform a meta-analysis to summarize existing survival data and reach a conclusion about the prognostic effect of MMP7 on the survival of GC patients.

Materials and Methods

Search strategy

A comprehensive search of electronic databases was completed from 1988 (when MMP7 was first introduced [41]) to October 29, 2014 in order to find clinical studies assessing the prognostic significance of MMP7 in GC. Databases searched included Medline, Embase, Web of Science, Google scholar, ProQuest (for dissertations) and Scopus. Key words were “matrix metalloproteinase 7”, “MMP-7” OR “matrilysin” AND ‘‘gastric” OR “stomach” AND “tumors”, ‘‘cancer”, ‘‘carcinoma”, ‘‘neoplasms” OR “CA” AND ‘‘survival,” ‘‘prognostic” or ‘‘prognosis.” The references cited in found full texts were scrutinized to find any additional studies not indexed in the databases that were searched. The language of the paper was not a matter of restriction for our search. We registered our review protocol in PROSPERO database ( which can be accessed using registry number CRD42014013770 (S1 File).

Study selection

Records in the primary search were assessed for relevance. The abstracts of relevant records were further screened to select appropriate items for full-text retrieval. Peer-reviewed and published studies addressing the association of MMP7 expression level (in blood or in resected gastric tumor specimen of histopathologically-confirmed gastric cancer patients) with patient survival were selected. A minimum of five years of follow-up was needed. Exclusion criteria were: in vitro and experimental studies, clinical cross-sectional studies, studies encompassing more than one type of cancer with no classified data, review articles, letters, editorials, conference abstracts and studies lacking the least requisite data to extract intended survival parameters for meta-analysis according to Palmar and Tierney [42,43]. When the results of the same cohort of patients were reported in more than one paper, the most informative study with the largest sample size was included. Eligible papers underwent quality appraisal and finally approved studies were included in the meta-analysis. We assumed no language restriction for study selection.

Quality assessment

Two independent reviewers (SSJ and NH) scored the quality of selected papers using the critical appraisal skills program (CASP) cohort study quality assessment checklist ( Then, they discussed their findings to reach a consensus on the final score of each paper. This checklist comprised 12 questions in three main parts (the validity of study, results and local implementation of results), and each appraised study received a score between 0 and 12. The quality scores achieved were used for subgroup analysis to check whether the quality of papers had an effect on the meta-analysis outcome or not.

Data extraction

Already prepared data tables were used to extract the necessary information. Two reviewers (SSJ and NH) performed the extraction process independently and the following discrepancies were resolved upon consensus or following the third reviewer’s (KZ) decision. When necessary, we contacted the corresponding authors for the required data. The parameters extracted included: the first author of the study, year of publication, source country of the patients, ethnicities, sample size, specimen assessed for MMP7 expression level, method of quantitative assessment, scoring system for MMP7 measurements and cut-off values used, histological grade and stage of the patients, positive expression rate of MMP7, other clinicopathological parameters reported, follow-up duration and survival parameters (Hazard ratio (HR) and its 95% confidence interval (CI), overall survival (OS) rate, disease-specific survival (DSS) rate, peritoneal recurrence free survival (PRFS) rate, relative hazard (RH) and its 95% CI, Log-Rank test indices, and Kaplan-Meier survival curves). The intended main outcome measurement for meta-analysis was extracted using already described approaches [4244]. The directly reported HR and its 95% CI were considered the most precise data. If not available, we tried to calculate it from Observed-Expected (O-E) event data in either group. Otherwise, we extracted the number of patients at risk, count of events and Log-Rank test indices to approximate HR and CI. Finally, if no informative numerical statistics were provided, survival curves were used. We exploited GetData Graph Digitizer software version ( in order to handle the curves as precisely as possible. This was to tackle the inter-reader variability limitation of this approach [44]. Uniform censoring throughout the follow-up period was assumed and previously developed methods [42,43] were used to calculate the censored number of cases in each time interval of survival curve data extraction.

Statistical methods

We used hazard ratios and the 95% CIs as the summary statistics for aggregated survival data, as already suggested [42] and the odds ratio (OR) and corresponding 95% CI to report aggregated association strength of MMP7 expression and other clinicopathological parameters. HR and OR above 1 were assumed to indicate poor prognosis and a positive association, respectively, provided that the 95% CIs did not overlap one.

Heterogeneity analysis was accomplished using both qualitative chi-square-based Q statistics and the quantitative metric I2 test (the number of studies analyzed does not affect the latter test) [45,46]. A P value < 0.05 for the Q statistics or I2 >50% indicated significant heterogeneity, necessitating a random-effect model for aggregated analysis. Otherwise, a fixed-effect model was used. I2 ≤ 50% would indicate a negligible quantitative degree of total variation among studies [45]. Sensitivity analysis was performed by the successive omission of individual studies to assess the integrity of summary results. Subgroup analysis based on intended parameters such as source country, measurement method and scoring system was performed to condition each parameter effect on summary results. Begg`s funnel plot and Egger`s test was exploited for potential publication bias assessment; an Egger`s test P value <0.10 would be interpreted as statistically significant [47].

We used Microsoft Excel 2013 for survival curves extracted data handling and Stata/SE version 11.1 software (Stata Corp LP, TX 77845, USA) for the rest of the analytical process.


Literature information

The initial search identified seventy four potentially relevant titles. By further reviewing the screening results, the reviewers determined twelve studies to be of acceptable relevance and format for the retrieval of full text. Among the selected papers, two [48,49] were excluded due to the lack of sufficient survival data and one [50] for data duplication; nine studies [3,27,3339] met the eligibility criteria and requisite quality (Fig 1) and were included in analysis of the prognostic value of MMP7 in GC as well as its association with clinicopathological parameters.

Fig 1. Flow diagram for study selection process.

The figure demonstrates how finally included studies were selected from primary search records.

Study characteristics

Table 1 shows the main characteristics of the selected eligible studies. In the table, studies are categorized by source of the samples used for MMP7 assay. A total of 1208 patients (ranging 42–264 for each study) were included in the analysis. The mean age of the patients was 64.26 (individual study mean range of 53.5–67); also, 62.11% of them were male, 38.86% had TNM early stage/stage 1 or 2 and 35.09% had well differentiated grade GC. Six studies were of an Asian population [3,33,3639] and the other three were from Finland [27], Brazil [35] and Spain [34]. IHC was used to assess MMP7 expression in five studies [3,27,33,35,38] and the other four exploited either a serum enzyme-linked immunosorbent assay (ELISA) [34,39] or real-time polymerase chain reaction (rt-PCR) [36,37]. Among the studies that used IHC, one study included invasive front of the tumor bulk in tissue sampling [38] and the others obtained specimens from random parts of the tumor [3,27,33,35]. Studies did not mention any control for IHC positivity. The MMP7 positive expression rate was 45–48.5% in four studies using IHC [3,27,33,35], and 66.7% in one using rt-PCR to detect MMP7 in different parts of gastric tumor specimens [36]; IHC-positive staining was up to 74.1% in the study assessing tumor invasive front [38]. The paper using rt-PCR to detect MMP7 mRNA in peritoneal lavage [37] reported a positive expression rate of 27%. The mean serum MMP7 concentration was reported to be 3.27 and 7.2 ng/ml in two studies using ELISA [34,39]. A 10–50% range of positive staining was used as the “high expression group” delineation cut-off point in IHC studies; it was 3.46 and 4.5 ng/ml of the serum MMP7 concentration in ELISA studies. Seven studies reported OS rate [3,27,3336,39], one reported DSS rate [38] and the other reported PRFS rate [37]. Five out of nine included studies that finally concluded the poor prognostic effect of MMP7 in GC [27,33,3638], while the other four did not reach such a conclusion [3,34,35,39]. A multivariate hazard ratio (HR) and its 95% confidence interval (CI) could be obtained for five studies (directly mentioned in the paper or sent by the corresponding author)[27,33,34,37,38]; just one of them reported direct univariate HR as well and we included multivariate HR for pooled analysis in this case [34]. For the rest of the studies [3,35,36,39], univariate HR was indirectly estimated from given Kaplan-Meier survival curves using previously described methods[4244]. Studies achieved a score of 8–10 out of 12 in quality appraisal. Comprehensive confounding factor consideration and applicability of study results to the local population were items that most of studies did not meet.

Summary Hazard Ratio

Directly obtained multivariate HRs of five studies and indirectly estimated univariate HRs of four studies were included for aggregated survival analysis. Our meta-analysis indicated the significant poor prognostic effect of MMP7 in GC patients with insignificant heterogeneity (pooled HR = 2.01, 95% CI = 1.62–2.50, Z = 6.32, P < 0.001, Fixed effect; Q = 10.948 on 8 degrees of freedom, P = 0.205, estimate of between studies variance = 0.043). Pooled HR and its 95% CI forest plot are depicted in Fig 2. Sensitivity analysis was performed by the successive omission of each study from aggregated survival meta-analyses to examine the influence of each individual study on the pooled HR. Fig 3 shows that none of the estimated pooled HRs corresponding to the omission of each study was outside the 95% CI of the HR estimated from all studies in the overall, implying that no individual study was dominant in the pooled results.

Fig 2. Forrest plot of overall hazard ratio estimate for MMP7 impact on GC survival.

The middle point of the diamond represents the pooled HR and its left and right corners represent 95% CI. Horizontal lines belong to individual studies; the middle point and line length represent the corresponding study`s extracted HR and 95% CI. The area of box tagged with each line represents the individual study`s weight of contribution to the meta-analysis.

Fig 3. Diagram for each study effect on pooled HR of MMP7 impact on GC survival.

The diagram illustrates sensitivity analysis results. Three bold vertical lines indicate pooled HR and it 95% CI when all studies included. Each dotted horizontal line belongs to a separate meta-analysis (with fixed effect model) when each respective study is omitted. The middle circle tagged represents corresponding pooled HR and two sides broken lines delineate its 95% CI.

Neither Begg`s nor Egger`s tests showed significant publication bias for the studies included for summary analysis (Begg`s test Z = 0.21, P = 0.83; Egger`s test t = 0.21, P = 0.84). Begg`s publication bias funnel plot is illustrated in Fig 4. The figure does not show apparent asymmetry.

Fig 4. Begg`s funnel plot for publication bias in MMP7 impact on GC survival meta-analysis.

Pseudo 95% Confidence limits are depicted and each hollow circle represents an individual study included in meta-analysis.

Subgroup analysis

Pooled HR and heterogeneity tests results for overall data and data restricted to different subgroups are given in Table 2. Although none of the subgroups had significantly different pooled HR compared to overall or counterpart subgroups, the numerical values of pooled HRs restricted to Asian populations, tissue sample MMP7 assay and 50% IHC cut-off point studies were notably higher than the pooled HRs restricted to the non-Asian populations, serum sample MMP7 assay and <50% IHC cut-off point data, respectively. Only two studies used serum MMP7 and their pooled HR was not statistically significant. The pooled HR limited to studies reporting multivariate analysis was not notably different from the overall summary HR. Restricting analysis to studies with mainly high stage patients (when more than two-third of study cases were high stage) revealed a slightly lower pooled HR. The overall HR was comparable with HR restricted to studies reporting the OS rate. According to quality appraisal, the included studies had acceptable and comparable quality scores; thereafter, no categorization for quality score was considered to make a subgroup analysis.

Table 2. Subgroup meta-analysis results for MMP7 impact on Gastric Cancer survival.

Association of MMP7 with Clinicopathological Parameters

Table 3 shows the meta-analysis results. The elevated expression of MMP7 was significantly associated with more aggressive tumor characteristics such as deeper invasion (pooled OR = 3.20, P = 0.026; fixed effect), higher TNM stage (pooled OR = 3.67,P<0.001; fixed effect), lymph node metastasis (pooled OR = 2.84,P<0.001; random effect), lymphatic vessels infiltration (pooled OR = 2.39, P = 0.024; fixed effect), blood vessels infiltration (pooled OR = 2.03, P = 0.026; fixed effect) and distant metastasis (pooled OR = 3.68,P<0.001; fixed effect). No association was detected between MMP7 expression and tumor size, histologic grade, age and sex (Table 3).

Table 3. Meta-analysis of MMP7 overexpression association with clinicopathological parameters in included studies.


For the first time, to the best of our knowledge, this meta-analysis on 1208 patients' pooled data indicated the MMP7 expression level to be significantly associated with poor survival in GC patients; the high expression group pooled probability of death was estimated to be almost twice that of the low expression group. A fixed effect was used to report pooled results regarding the insignificant heterogeneity test. Sensitivity analysis showed that no individual study significantly dominated the pooled result. No publication bias was detected according to Begg's and Egger's tests. The clinicopathological data analysis in the same studies was meant to assess whether it is consistent with survival data; in support, it was shown that MMP7 level is significantly associated with aggressive tumor characteristics such as invasion depth, TNM stage, and distant metastasis.

By degrading ECM proteins and regulating the activity of other biomolecules in the body, MMPs mediate many processes such as cell migration, differentiation, proliferation, apoptosis, inflammatory reactions, and angiogenesis, both in physiological conditions (embryogenesis) and pathological disorders (cancer)[4,40]. The enzymes can affect pivotal steps of cancer biology such as growth, survival, angiogenesis and invasion [4]. MMP7 is a key member of this family that inherits the same properties as well as other specific features [1921,26,51,52], which suggests that it is an affective biomolecule in tumorigenesis and cancer progression. Subsequent studies provided clinical evidence for these molecular findings [23,2732]. Similar to other cancers, many authors aimed to assess the MMP7 prognostic role in GC survival. Some of them concluded that MMP7 is a poor prognostic factor for GC [27,33,3638], while others[3,34,35,39] did not. Our meta-analysis concluded that MMP7 is a poor prognostic factor of GC survival.

A recent meta-analysis reported summary odds ratio data of MMP7 association with GC pathological indices [40]. The authors revealed that high MMP7 expression was associated with aggressive tumor phenotypes such as TNM stage, depth of tumor invasion, lymph node and distant metastasis. They detected no association between MMP7 and histological grade. Accordingly, in our included survival studies, we found that high MMP7 level is correlated with invasion depth, lymph node metastasis, distant metastasis and TNM staging. We did not find any significant association between MMP7 and grade either. The mentioned meta-analysis [40] did not include follow up survival data and did not address the inconsistency in results of studies that have investigated prognostic effect of MMP7 in gastric cancer.

Our subgroup analysis showed that pooled HR of Asian patients, tumor tissue MMP7 and 50% IHC cut-off point categories were numerically higher than the categories of non-Asian patients, serum MMP7 and <50% IHC cut-off point studies, respectively; however, they were all statistically comparable (as already shown by the insignificant heterogeneity test). Genetic background as well as environmental factors vary in different regions. This leads to tumor generation with different biological behavior [7]. In addition, tumor site differs between eastern and western countries, with consequent differences in tumor behavior and prognosis [53]. These could be reasons for the numerical differences detected between the pooled HR of Asian and non-Asian patients.

In patients with solid tumors, serum level of MMP7 does not seem to correlate strongly with tumor tissue level of MMP7. In breast, colorectal and ovarian cancers, no significant correlation was detected and just a weak correlation was reported in gastric cancer [5457]. Accordingly, we found different results for prognostic effect of tissue and serum MMP7 level in gastric cancer. Although the summary estimates for association between tumor tissue MMP7 and prognosis of gastric cancer was significant, the pooled HR of two studies that used serum MMP7 was not statistically significant. Further studies are needed to assess association of serum MMP7 and the prognosis of gastric cancer.

Considering the higher values of MMP7-positive IHC staining as a cut-off to delineate the "high expression group" further discriminated between the prognosis of low and high expression groups and enhanced the specificity of MMP7 as a prognostic biomarker. Summary HR of studies with multivariate analysis data which encompasses about two-thirds of the included cases denoted a significant poor prognostic role of MMP7; this implies that the extracted summary prognostic effect of MMP7 in gastric cancer could be considered independent of other known prognostic factors (e.g. stage). Seven studies reported OS [3,27,3336,39], one reported DSS [38] and the other reported PRFS [37]. Pooling data of three types of survival did not produce significant heterogeneity. In addition, pooled HR restricted to OS-reporting studies did not differ notably from overall HR.

The invasive front of tumor specimens showed a higher MMP7 expression rate compared to other parts of the tumor [38]. In addition, the type of antibody used and the degree of dilution could reveal different results [7]. Therefore, sampling site and IHC method (regarding type of antibody used and degree of dilution) should be standardized to address such measurement bias when introducing an IHC-based biomarker.

This study reached a homogeneous significant conclusion about the poor prognostic effect of MMP7 in GC patients' survival. This finding was supported by the association of MMP7 with aggressive tumor clinicopathological characteristics. Homogeneity further strengthened the meta-analysis and indicated that the analyzed data were similar enough to be pooled and that the summary results would be trustworthy. In general, HR > 2 is considered notably predictive [58]. Extracted prognostic effect of MMP7 in gastric cancer merits notice from two clinical aspects. It introduces MMP7 as a potential target for molecular anti-cancer therapy in gastric cancer. Current literature considers MMP7 a validated target for anticancer drugs [59]. Batimastat and Marimastat are broad spectrum MMP antagonists that target MMP7 along with some other MMPs [60,61]. There is experimental evidence for anti-proliferative and anti-metastatic effects of Batimastat [62]. Phase II and III clinical trials as well as observational studies demonstrate benefit of Marimastat administration in gastrointestinal malignancies [60,61,63,64]. Also an experimental study reports inhibitory effect of MMP7-specific antisense oligonucleotide on peritoneal dissemination in human GC [65]. In a certain type of cancer, a selection of MMPs with established poor prognostic effects could be a potentially good combination for targeted therapy.

In addition to be a target for GC therapy, MMP7 appears to be a good candidate for molecular staging of GC in the clinic to improve conventional clinical staging. It could help to characterize the patients more precisely and categorize them in appropriate therapeutic group. Further investigations are required to reach a consensus on appropriate method and cut-off values. In the biological milieu of tumor cells, the vast number of biomarkers were found to interact. Thereafter, an appropriately selected combination of biomarkers, instead of one item alone, could be considered a molecular signature of a tumor to complement the clinical staging.

Overall, the number of patients included in this meta-analysis as well as sample size in some of individual papers were small. This could limit the strength of our findings. However, pooled HR restricted to larger studies did not significantly differ from the overall pooled HR. Only five papers directly reported HRs; we extracted HR from survival curves in other studies. This method of data extraction is prone to bias [42]; we tried to limit this as much as possible (e.g. using a graphical curve reader software to read curves and choosing appropriate time intervals). Most studies did not consider a comprehensive profile of confounding factors to extract a very independent and pure prognostic effect of MMP7. Our search strategy targeted peer-reviewed published papers. Unpublished data and conference presentations were not included.

Future cohort studies with a larger number of patients would produce more robust results. In addition, well-designed multivariate survival analyses (such as multivariate Cox proportional hazard model) conditioning a comprehensive number of probable confounding variables are recommended to extract the independent pure prognostic effect of MMP7. A selection of appropriate molecular candidates along with MMP7 in combination with clinical data, could be assessed for outcome-predicting capability in GC. Such a study needs a very large sample size and requires collaboration of different groups and cancer centers.

For the first time, to the best of our knowledge, this meta-analysis on the statistically homogenous data from 1208 patients concluded that MMP7 is a poor prognostic factor for the survival of GC patients. Additionally, its overexpression was correlated with more advanced clinicopathological features. MMP7 alone, or more appropriately in combination with other biomarkers, could be considered a prognostic biomarker in the clinic to predict the outcome of GC patients, especially in Asian populations.

Excluded Papers

  1. Kim JH, Eom DY, Kim CW, Choi NK, Kwak JH, et al. (2011) Expression of E-cadherin, β-catenin, Cdx2 and MMP7 in pT2 and N1/N2 Gastric Cancer: Relationship with Tumor Recurrence within 2-Year Period. Journal of the Korean Surgical Society 80: 29–35 (Ref #35).
    Reason for exclusion: The paper does not present sufficient survival data (Directly reported HR and its 95% CI, Observed-Expected (O-E) event data in either group, count of events and Log-Rank test indices or survival curves) and we could not generate HR.
  2. Kubben FJ, Sier CF, van Duijn W, Griffioen G, Hanemaaijer R, et al. (2006) Matrix metalloproteinase-2 is a consistent prognostic factor in gastric cancer. Br J Cancer 94: 1035–1040 (Ref #36).
    Reason for exclusion: The paper does not present sufficient survival data (Directly reported HR and its 95% CI, Observed-Expected (O-E) event data in either group, count of events and Log-Rank test indices or survival curves) and we could not generate HR.
  3. Yonemura Y, Endou Y, Fujita H, Fushida S, Bandou E, et al. (2000) Role of MMP-7 in the formation of peritoneal dissemination in gastric cancer. Gastric Cancer 3: 63–70 (Ref # 40).
    Reason for exclusion: Gastric cancer patients included in this study are part of the patients included in the more recent paper "Ajisaka H, Yonemura Y, Miwa K (2004) Correlation of lymph node metastases and expression of matrix metalloproteinase-7 in patients with gastric cancer. Hepatogastroenterology 51: 900–905 (Ref # 33) ". The later paper used the same material with a larger sample size. Therefore, we decide to use the recent paper (Ref # 33) and excluded the first report (Ref # 40).

Supporting Information

S1 PRISMA Checklist. PRISMA checklist for systematic review studies.

The file includes completed checklist for this study appraisal.


S1 File. Registered review protocol.

The file is a print of review protocol registered in PROSPERO review registration database.


Author Contributions

Conceived and designed the experiments: SS AA RH. Performed the experiments: SS KZ NH. Analyzed the data: SS KZ SN. Contributed reagents/materials/analysis tools: SN AA NH RH. Wrote the paper: SS KZ SN AA NH RH.


  1. 1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM (2010) Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer 127: 2893–2917. pmid:21351269
  2. 2. Kim BS, Cho SW, Min SK, Lee BH (2011) Differences in prognostic factors between early and advanced gastric cancer. Hepatogastroenterology 58: 1032–1040. pmid:21830438
  3. 3. Sawada T, Yashiro M, Sentani K, Oue N, Yasui W, Miyazaki K, et al. (2015) New molecular staging with G-factor supplements TNM classification in gastric cancer: a multicenter collaborative research by the Japan Society for Gastroenterological Carcinogenesis G-Project committee. Gastric Cancer 18: 119–128 (Epub 2014 Feb 1). pmid:24488015
  4. 4. Hadler-Olsen E, Winberg JO, Uhlin-Hansen L (2013) Matrix metalloproteinases in cancer: their value as diagnostic and prognostic markers and therapeutic targets. Tumour Biol 34: 2041–2051. pmid:23681802
  5. 5. Brinckerhoff CE, Matrisian LM (2002) Matrix metalloproteinases: a tail of a frog that became a prince. Nat Rev Mol Cell Biol 3: 207–214. pmid:11994741
  6. 6. Butler GS, Overall CM (2009) Proteomic identification of multitasking proteins in unexpected locations complicates drug targeting. Nat Rev Drug Discov 8: 935–948. pmid:19949400
  7. 7. Shen W, Xi H, Wei B, Chen L (2014) The prognostic role of matrix metalloproteinase 2 in gastric cancer: a systematic review with meta-analysis. J Cancer Res Clin Oncol 140: 1003–1009. pmid:24610446
  8. 8. Zhang QW, Liu L, Chen R, Wei YQ, Li P, Shi HS, et al. (2012) Matrix metalloproteinase-9 as a prognostic factor in gastric cancer: a meta-analysis. Asian Pac J Cancer Prev 13: 2903–2908. pmid:22938481
  9. 9. Jezierska A, Motyl T (2009) Matrix metalloproteinase-2 involvement in breast cancer progression: a mini-review. Med Sci Monit 15: Ra32-40.
  10. 10. Song J, Su H, Zhou YY, Guo LL (2013) Prognostic value of matrix metalloproteinase 9 expression in breast cancer patients: a meta-analysis. Asian Pac J Cancer Prev 14: 1615–1621. pmid:23679245
  11. 11. Peng WJ, Zhang JQ, Wang BX, Pan HF, Lu MM, Wang J (2012) Prognostic value of matrix metalloproteinase 9 expression in patients with non-small cell lung cancer. Clin Chim Acta 413: 1121–1126. pmid:22465234
  12. 12. Qian Q, Wang Q, Zhan P, Peng L, Wei SZ, Shi Y, et al. (2010) The role of matrix metalloproteinase 2 on the survival of patients with non-small cell lung cancer: a systematic review with meta-analysis. Cancer Invest 28: 661–669. pmid:20394501
  13. 13. Li CY, Yuan P, Lin SS, Song CF, Guan WY, Yuan L, et al. (2013) Matrix metalloproteinase 9 expression and prognosis in colorectal cancer: a meta-analysis. Tumour Biol 34: 735–741. pmid:23269605
  14. 14. van der Jagt MF, Wobbes T, Strobbe LJ, Sweep FC, Span PN (2010) Metalloproteinases and their regulators in colorectal cancer. J Surg Oncol 101: 259–269. pmid:20039272
  15. 15. Gadducci A, Cosio S, Tana R, Genazzani AR (2009) Serum and tissue biomarkers as predictive and prognostic variables in epithelial ovarian cancer. Crit Rev Oncol Hematol 69: 12–27. pmid:18595727
  16. 16. Szarvas T, vom Dorp F, Ergun S, Rubben H (2011) Matrix metalloproteinases and their clinical relevance in urinary bladder cancer. Nat Rev Urol 8: 241–254. pmid:21487384
  17. 17. Matrisian LM (1992) The matrix-degrading metalloproteinases. Bioessays 14: 455–463. pmid:1445287
  18. 18. Miyazaki K, Hattori Y, Umenishi F, Yasumitsu H, Umeda M (1990) Purification and characterization of extracellular matrix-degrading metalloproteinase, matrin (pump-1), secreted from human rectal carcinoma cell line. Cancer Res 50: 7758–7764. pmid:2253219
  19. 19. Parsons SL, Watson SA, Brown PD, Collins HM, Steele RJ (1997) Matrix metalloproteinases. Br J Surg 84: 160–166. pmid:9052425
  20. 20. Imai K, Yokohama Y, Nakanishi I, Ohuchi E, Fujii Y, Nakai N, et al. (1995) Matrix metalloproteinase 7 (matrilysin) from human rectal carcinoma cells. Activation of the precursor, interaction with other matrix metalloproteinases and enzymic properties. J Biol Chem 270: 6691–6697. pmid:7896811
  21. 21. Adachi Y, Yamamoto H, Itoh F, Arimura Y, Nishi M, Endo T, et al. (2001) Clinicopathologic and prognostic significance of matrilysin expression at the invasive front in human colorectal cancers. Int J Cancer 95: 290–294. pmid:11494227
  22. 22. Pajouh MS, Nagle RB, Breathnach R, Finch JS, Brawer MK, Bowden GT (1991) Expression of metalloproteinase genes in human prostate cancer. J Cancer Res Clin Oncol 117: 144–150. pmid:1848860
  23. 23. Yamamoto H, Adachi Y, Itoh F, Iku S, Matsuno K, Kusano M, et al. (1999) Association of matrilysin expression with recurrence and poor prognosis in human esophageal squamous cell carcinoma. Cancer Res 59: 3313–3316. pmid:10416584
  24. 24. Ii M, Yamamoto H, Adachi Y, Maruyama Y, Shinomura Y (2006) Role of matrix metalloproteinase-7 (matrilysin) in human cancer invasion, apoptosis, growth, and angiogenesis. Exp Biol Med (Maywood) 231: 20–27. pmid:16380641
  25. 25. Wielockx B, Libert C, Wilson C (2004) Matrilysin (matrix metalloproteinase-7): a new promising drug target in cancer and inflammation? Cytokine Growth Factor Rev 15: 111–115. pmid:15110795
  26. 26. Ii M, Yamamoto H, Adachi Y, Maruyama Y, Shinomura Y (2006) Role of matrix metalloproteinase-7 (matrilysin) in human cancer invasion, apoptosis, growth, and angiogenesis. Exp Biol Med 231: 20–27. pmid:16380641
  27. 27. Koskensalo S, Mrena J, Wiksten JP, Nordling S, Kokkola A, Hagstrom J, et al. (2010) MMP-7 overexpression is an independent prognostic marker in gastric cancer. Tumour Biol 31: 149–155. pmid:20300917
  28. 28. Liu D, Nakano J, Ishikawa S, Yokomise H, Ueno M, Kadota K, et al. (2007) Overexpression of matrix metalloproteinase-7 (MMP-7) correlates with tumor proliferation, and a poor prognosis in non-small cell lung cancer. Lung Cancer 58: 384–391. pmid:17728005
  29. 29. Koskensalo S, Louhimo J, Nordling S, Hagstrom J, Haglund C (2011) MMP-7 as a prognostic marker in colorectal cancer. Tumour Biol 32: 259–264. pmid:21207220
  30. 30. Beeghly-Fadiel A, Shu XO, Long J, Li C, Cai Q, Cai H, et al. (2009) Genetic polymorphisms in the MMP-7 gene and breast cancer survival. Int J Cancer 124: 208–214. pmid:18798254
  31. 31. Szarvas T, Becker M, Vom Dorp F, Meschede J, Scherag A, Bankfalvi A, et al. (2011) Elevated serum matrix metalloproteinase 7 levels predict poor prognosis after radical prostatectomy. Int J Cancer 128: 1486–1492. pmid:20473942
  32. 32. Szarvas T, Jager T, Becker M, Tschirdewahn S, Niedworok C, Kovalszky I, et al. (2011) Validation of circulating MMP-7 level as an independent prognostic marker of poor survival in urinary bladder cancer. Pathol Oncol Res 17: 325–332. pmid:21253919
  33. 33. Ajisaka H, Yonemura Y, Miwa K (2004) Correlation of lymph node metastases and expression of matrix metalloproteinase-7 in patients with gastric cancer. Hepatogastroenterology 51: 900–905. pmid:15143943
  34. 34. Blanco-Calvo M, Tarrio N, Reboredo M, Haz-Conde M, Garcia J, Quindos M, et al. (2014) Circulating levels of GDF15, MMP7 and miR-200c as a poor prognostic signature in gastric cancer. Future Oncol 10: 1187–1202. pmid:24947260
  35. 35. Fanelli MF, Chinen LT, Begnami MD, Costa WL Jr., Fregnami JH, Soares FA, et al. (2012) The influence of transforming growth factor-alpha, cyclooxygenase-2, matrix metalloproteinase (MMP)-7, MMP-9 and CXCR4 proteins involved in epithelial-mesenchymal transition on overall survival of patients with gastric cancer. Histopathology 61: 153–161. pmid:22582975
  36. 36. Lee KH, Shin SJ, Kim KO, Kim MK, Hyun MS, Kim TN, et al. (2006) Relationship between E-cadherin, matrix metalloproteinase-7 gene expression and clinicopathological features in gastric carcinoma. Oncol Rep 16: 823–830. pmid:16969501
  37. 37. Li Z, Zhang D, Zhang H, Miao Z, Tang Y, Sun G, et al. (2014) Prediction of peritoneal recurrence by the mRNA level of CEA and MMP-7 in peritoneal lavage of gastric cancer patients. Tumour Biol 35: 3463–3470. pmid:24282089
  38. 38. Liu XP, Kawauchi S, Oga A, Tsushimi K, Tsushimi M, Furuya T, et al. (2002) Prognostic significance of matrix metalloproteinase-7 (MMP-7) expression at the invasive front in gastric carcinoma. Jpn J Cancer Res 93: 291–295. pmid:11927011
  39. 39. Yeh YC, Sheu BS, Cheng HC, Wang YL, Yang HB, Wu JJ (2010) Elevated serum matrix metalloproteinase-3 and -7 in H. pylori-related gastric cancer can be biomarkers correlating with a poor survival. Dig Dis Sci 55: 1649–1657. pmid:19690958
  40. 40. Long ZW, Wang JL, Wang YN (2014) Matrix metalloproteinase-7 mRNA and protein expression in gastric carcinoma: a meta-analysis. Tumour Biol 35: 11415–11426 pmid:25123263
  41. 41. Woessner JF Jr., Taplin CJ (1988) Purification and properties of a small latent matrix metalloproteinase of the rat uterus. J Biol Chem 263: 16918–16925. pmid:3182822
  42. 42. Parmar MK, Torri V, Stewart L (1998) Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 17: 2815–2834. pmid:9921604
  43. 43. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR (2007) Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 8: 16. pmid:17555582
  44. 44. Williamson PR, Smith CT, Hutton JL, Marson AG (2002) Aggregate data meta-analysis with time-to-event outcomes. Stat Med 21: 3337–3351. pmid:12407676
  45. 45. Higgins JP, Thompson SG (2002) Quantifying heterogeneity in a meta-analysis. Stat Med 21: 1539–1558. pmid:12111919
  46. 46. Higgins JP, Thompson SG, Deeks JJ, Altman DG (2003) Measuring inconsistency in meta-analyses. BMJ 327: 557–560. pmid:12958120
  47. 47. Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias in meta-analysis detected by a simple, graphical test. BMJ 315: 629–634. pmid:9310563
  48. 48. Kim JH, Eom DY, Kim CW, Choi NK, Kwak JH, Choi GM, et al. (2011) Expression of E-cadherin, β-catenin, Cdx2 and MMP7 in pT2 and N1/N2 Gastric Cancer: Relationship with Tumor Recurrence within 2-Year Period. Journal of the Korean Surgical Society 80: 29–35.
  49. 49. Kubben FJ, Sier CF, van Duijn W, Griffioen G, Hanemaaijer R, van de Velde CJ, et al. (2006) Matrix metalloproteinase-2 is a consistent prognostic factor in gastric cancer. Br J Cancer 94: 1035–1040. pmid:16538217
  50. 50. Yonemura Y, Endou Y, Fujita H, Fushida S, Bandou E, Taniguchi K, et al. (2000) Role of MMP-7 in the formation of peritoneal dissemination in gastric cancer. Gastric Cancer 3: 63–70. pmid:11984713
  51. 51. von Bredow DC, Nagle RB, Bowden GT, Cress AE (1997) Cleavage of beta 4 integrin by matrilysin. Exp Cell Res 236: 341–345. pmid:9344615
  52. 52. Wang WS, Chen PM, Wang HS, Liang WY, Su Y (2006) Matrix metalloproteinase-7 increases resistance to Fas-mediated apoptosis and is a poor prognostic factor of patients with colorectal carcinoma. Carcinogenesis 27: 1113–1120. pmid:16474169
  53. 53. Wen L, Chen XZ, Yang K, Chen ZX, Zhang B, Chen JP, et al. (2013) Prognostic value of cancer stem cell marker CD133 expression in gastric cancer: a systematic review. PLoS One 8: e59154. pmid:23533603
  54. 54. Gershtein ES, Korotkova EA, Shcherbakov AM, Prorokov VV, Golovkov DA, Kushlinskii NE (2007) Matrix metalloproteinases 7 and 9 and their types 1 and 4 tissue inhibitors in tumors and plasma of patients with colorectal cancer. Bull Exp Biol Med 143: 459–462. pmid:18214300
  55. 55. Gershtein ES, Levkina NV, Digayeva MA, Laktionov KP, Tereshkina IV, Kushlinsky NE (2010) Matrix metalloproteinases 2, 7, and 9 and tissue inhibitor of metalloproteinases-1 in tumors and serum of patients with ovarian neoplasms. Bull Exp Biol Med 149: 628–631. pmid:21165404
  56. 56. Gerstein ES, Sini L, Ryabov AB, Dvorova EK, Yurchenko AA, Stilidi IS, et al. (2009) Comparative enzyme immunoassay of matrix metalloproteinases-2, -7, -9 and their tissue inhibitor-2 in tumors and plasma of patients with gastric cancer. Bull Exp Biol Med 148: 899–902. pmid:21116502
  57. 57. Katunina AI, Gershtein ES, Ermilova VD, Tereshkina IV, Nazarenko AY, Tyleuova AA, et al. (2011) Matrix metalloproteinases 2, 7, and 9 in tumors and sera of patients with breast cancer. Bull Exp Biol Med 151: 359–362. pmid:22451887
  58. 58. Ferlay J, Parkin DM, Steliarova-Foucher E (2010) Estimates of cancer incidence and mortality in Europe in 2008. Eur J Cancer 46: 765–781. pmid:20116997
  59. 59. Overall CM, Kleifeld O (2006) Tumour microenvironment—opinion: validating matrix metalloproteinases as drug targets and anti-targets for cancer therapy. Nat Rev Cancer 6: 227–239. pmid:16498445
  60. 60. Chaudhary AK, Pandya S, Ghosh K, Nadkarni A (2013) Matrix metalloproteinase and its drug targets therapy in solid and hematological malignancies: an overview. Mutat Res 753: 7–23. pmid:23370482
  61. 61. Vihinen P, Kahari VM (2002) Matrix metalloproteinases in cancer: prognostic markers and therapeutic targets. Int J Cancer 99: 157–166. pmid:11979428
  62. 62. Davies B, Brown PD, East N, Crimmin MJ, Balkwill FR (1993) A synthetic matrix metalloproteinase inhibitor decreases tumor burden and prolongs survival of mice bearing human ovarian carcinoma xenografts. Cancer Res 53: 2087–2091. pmid:8347186
  63. 63. Tierney GM, Griffin NR, Stuart RC, Kasem H, Lynch KP, Lury JT, et al. (1999) A pilot study of the safety and effects of the matrix metalloproteinase inhibitor marimastat in gastric cancer. Eur J Cancer 35: 563–568. pmid:10492628
  64. 64. North H, King J, Morris DL (2000) Effect of marimastat on serum tumour markers in patients with colorectal cancer. Int J Surg Investig 2: 213–217. pmid:12678521
  65. 65. Yonemura Y, Endo Y, Fujita H, Kimura K, Sugiyama K, Momiyama N, et al. (2001) Inhibition of peritoneal dissemination in human gastric cancer by MMP-7-specific antisense oligonucleotide. J Exp Clin Cancer Res 20: 205–212. pmid:11484976