Overview of Methodological Quality of Systematic Reviews about Gastric Cancer Risk and Protective Factors

BACKGROUND AND OBJECTIVE
A comprehensive overall review of gastric cancer (GC) risk and protective factors is a high priority, so we conducted the present study.


METHODS
Systematic searches in common medical electronic databases along with reference tracking were conducted to include all kinds of systematic reviews (SRs) about GC risk and protective factors. Two authors independently selected studies, extracted data, and evaluated the methodological qualities and the quality of evidence using R-AMSTAR and GRADE approaches.


RESULTS
Beta- carotene below 20 mg/day, fruit, vegetables, non-fermented soy-foods, whole-grain, and dairy product were GC protective factors, while beta-carotene 20 mg/day or above, pickled vegetables, fermented soy-foods, processed meat 30 g/d or above, or salty foods, exposure to alcohol or smoking, occupational exposure to Pb, overweight and obesity, helicobacter pylori infection were GC risk factors. So we suggested screening and treating H. pylori infection, limiting the amount of food containing risk factors (processed meat consumption, beta-carotene, pickled vegetables, fermented soy-foods, salty foods, alcohol), stopping smoking, avoiding excessive weight gain, avoidance of Pb, and increasing the quantity of food containing protective components (fresh fruit and vegetables, non-fermented soy-foods, whole-grain, dairy products).


CONCLUSIONS
The conclusions and recommendations of our study were limited by including SRs with poor methodological bases and low quality of evidence, so that more research applying checklists about assessing the methodological qualities and reporting are needed for the future.


Introduction
Although the incidence and mortality rates of gastric cancer (GC) have decreased markedly in the past few decades, GC is the fourth most common cancer (accounted for 988,602, about 7.8% of all cancers) and the second leading cause of cancer-related death (accounted for 737,419, about 9.7% of all cancers) worldwide in 2008 (Jemal et al., 2002).As a result of trends in global aging and population growth, the potential incidence of GC for 2010 is estimated to achieve 1.1 million (Wang et al., 2010).So actions must be taken, in order to lower the chance of getting GC and the number of new GC cases.The best way is to avoid the risk factors associated with GC (any attribute, characteristic or exposure of an individual that increases the likelihood of developing GC) and increase GC protective factors (any attribute, characteristic or exposure of an individual that that prevents or reduces vulnerability for the development of GC).To prevent new GC from starting, risk factors and protective factors were needed to be found and recognized, and recommendations

Overview of Methodological Quality of Systematic Reviews about Gastric Cancer Risk and Protective Factors
Lun Li 1,2& , Xiang-Ji Ying 1,3& , Tian-Tian Sun 4 , Kang Yi 1,2 , Hong-Liang Tian 1,2 , Rao Sun 1 , Jin-Hui Tian 1 , Ke-Hu Yang 1,2 * should be made at the same time.There were guidelines on nutrition and physical activity for GC prevention that produced by American Cancer Society (Byers et al., 2002) or the Asia-Pacific Gastric Cancer Consensus Conference (Fock et al., 2008), but these guidelines were not scientific enough using the best evidence, or comprehensive enough using all available evidence.So we conducted this study to review all available evidence about the GC etiology based on SRs or meta-analyses to present all GC risk and protective factors and give some recommendations by the way.

Data source and study selection
The Cochrane Library, PubMed, Embase, ISI Web of Knowledge, China Academic Journal Network Publishing Database, and Chinese Scientific Journals Full text Database and Chinese Biomedical Literature Database were searched with ("stomach cancer*" OR "stomach neoplasm*" OR "stomach tumor*" OR "stomach adenocarcinoma" OR "gastric cancer*" OR "gastric neoplasm*" OR "gastric tumor*" OR "gastric adenocarcinoma") and ("meta analys*" OR "review*") without language, publication year and publication status restrictions.All searches were conducted in March 2010 and updated in April 2011 by two independent reviewers (Lun Li & Tiantian Sun).Additional citations were identified by checking references of included studies and existing reviews of this topic.
Factors that were related to GC from available SRs or meta-analyses were divided into four categories (risk factor, protective factor, non-significant factor, unclear factor).The risk factor is any attribute, characteristic or exposure that increases the likelihood of developing GC and the protective factor is any attribute, characteristic or exposure that prevents or reduces vulnerability in the development of GC.Those factors that were not significant in the GC developing were defined as non-significant factors, and those factors whose effects in GC developing were unclear were called unclear factors.When there were more than one SRs or meta-analyses that focused on the same topic, we compared the differences among them and included the most comprehensive one.
Two reviewers (Lun Li & Tiantian Sun) independently selected studies according to predetermined inclusion criteria by screening titles and abstracts identified through all searches.If both reviewers believed that the abstracts were potentially relevant, we screened the full-text articles independently to determine whether the study was were eligible for inclusion.

Data collection and quality assessment
Two reviewers (Lun Li & Tiantian Sun) independently extracted data, assessed methodological quality, and evaluated the quality of evidence.Disagreements were resolved by consultations with a third reviewer (Kehu Yang).The data extraction form summarized key characteristics, including information on participants, exposures, outcomes, conclusions, and the quality assessment items.
The methodological qualities of included SRs were evaluated using assessment of multiple systematic reviews (AMSTAR), because it has good content validity, wide acceptance, recognized reliability and reproducibility (Shea et al., 2007;Shea et al., 2007;Shea et al., 2009;Kung et al., 2010).Revised Assessment of Multiple Systematic Reviews (R-AMSTAR) was chosen in our study, as it could produce quantifiable assessments of SR quality (Kung et al., 2010).
For the evaluation of quality of evidence, we used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (Guyatt et al., 2010).The GRADE approach specifies four levels of quality: high, moderate, low, and very low after evaluating five factors (study quality, consistency, directness, precision, and reporting bias) which may lead to its downgrading and three factors (large effects, all plausible residual confounding and dose-response gradient) which can lead to upgrading quality of evidence (Nasser and Fedorowicz, 2011).

Main results from SRs or meta-analyses (Table 1)
Available evidence showed that the relationships between diabetes, NSAIDs (aspirin), preserved fish, preserved vegetable, smoked food, green tea, red meat, beer intake, glycemic load, vitamin C, ham, sausage and GC were inconsistent.And there were not significant associations between occupational exposure to Cr and chrysotile fiber, dupA H. pylori positive strains, glycemic index, folate, coffee, fish, selenium or statins, omega-3 fatty acids, vitamin E, aspirin, onion leaf and GC.
The methodological quality of included SRs or metaanalyses (Table 2).
Of 52 small items from the R-AMSTAR (Kung et al., 2010), seven small items were not conducted in all included studies, seventeen small items were conducted in more than half of all included studies (five small items were conducted in more than 75% of all included studies), twenty-seven small items were conducted in less than half of all included studies (eleven small items were conducted in less than 25% of all included studies).
For item 1, none of these 55 SRs or meta-analyses conducted 'a priori' design, as all of them were published in regular journals.96.36% of all studies stated inclusion criteria and 65.45% raised their research questions using PICO format.
For item 2, 27.27% studies did duplicate study selection and data extraction and 56.36% showed that they conducted none of three small items in item 2. All these three small items were conducted in less than half of all included studies (32.73%).
For item 3, which was operationalized into five small items, none of all studies conducted all and only five studies conducted none.52.73% searched at least two electronic sources, 49.09% told the search time, 83.64% mentioned search strategy or key words they used to search, 67.27% also searched by reviewing the references in the studies found or consulting current contents,   The total number of events is less than 300 reviews, textbooks, specialized registers, or experts in the particular field, none hand-searched relevant journals.
For item 4, 89.09% studies did not use the status of publication (i.e.grey literature) as an inclusion criterion, and no study conducted all four small items operationalized from item 4. 3.64% stated that they searched for reports regardless of their publication type, 1.82% stated they excluded any reports based on their publication status, none mentioned "Non-English papers were translated", and 9.09% said no language restriction.
For item 5, 9.09% did not provide a list of studies (included and excluded), and none conducted all four small items operationalized from item 5. 70.91% gave a table of included studies, 1.82% mentioned excluded studies in a supplemental source, 36.36%stated the reason for exclusion of the seriously considered studies, and 90.91% supplied a table or reference list of included or excluded studies.
For item 6, 16.36% provided the characteristics of the included studies, and none conducted all three small items operationalized from item 6. 83.64% provided a table that presented the original data about PICO, 70.91% provided the ranges of relevant characteristics in the studies analyzed, and 83.64% provided complete information.
For item 7, 81.82% did not assess the scientific quality of the included studies, and none conducted all four small items operationalized from item 7. 18.18% provided 'a priori' methods of assessment, 12.73% provided the assessment results, and none discussed level of evidence and ranked quality of evidence.
For item 8, 92.73% did not use the scientific quality of the included studies in formulating conclusions, and 1.82% conducted all four small items operationalized from item 8. 7.27% considered the results of the methodological rigor and scientific quality, and 7.27% stated the results of the methodological rigor and scientific quality.
For item 9, 25.45% did not tell how to combine the findings of studies, and none conducted all five small items operationalized from item 9. None stated criteria that the studies analyzed were similar enough to be pooled, 50.91% did a test to assess their homogeneity, 67.27% recognized heterogeneity, and 30.91% supplied the methods to deal with high heterogeneity.
For item 10, 47.27% did not assess the likelihood of publication bias, and 1.82% conducted all three small items operationalized from item 8. 56.36% recognized publication bias, 43.64% assessed publication bias using graphical aids, and 38.18% used statistical tests.

Discussion
Diet and lifestyle are thought to be involved in the development of GC.The American Cancer Society recommends a diet that is high in fresh fruit, fresh vegetables, and whole grain foods and low in processed food and red meat.This is similar to our results, namely intake of food containing protective factors (fruit, vegetables, soybean products, non-fermented soy-foods, whole-grain, and dairy product) could prevent GC.Certainly, some food that contains chemical carcinogenic compounds may be involved in increasing the risks of developing GC, such as nitrites and nitrates, which could be transformed to carcinogenic compounds nitrosamines inside the human body.In our study, intake of food containing risk factors (pickled vegetables, fermented soy-foods, processed meat 30 g/d or above, and bacon, or salted foods), might be associated with increased GC risk.Recent study has showed that increased consumption of vegetables, fruit and vitamin C and a decrease in salt consumption could result the decline of GC incidence in Poland for about forty years (Jarosz et al., 2011).So in the future, increasing intake of food containing protective factors, and limiting or avoiding intake of food containing risk factors might help to prevent GC.As a result, diet that contains more protective factors and less or no risk factors should be recommended for GC prevention.Lifestyle exposure to alcohol and smoking were associated with increased GC risks (Tredaniel et al., 1997;Corrao et al., 1999;Bagnardi et al., 2001;Bagnardi et al., 2001;Liu and Wang, 2002;Zeka et al., 2003;Corrao et al., 2004;Inoue et al., 2005;Nishino et al., 2006;Gandini et al., 2008;Ladeiras-Lopes et al., 2008;Shimazu et al., 2008;La Torre et al., 2009;Li et al., 2011;Liu et al., 2009;Tramacere et al., 2011;Tramacere et al., 2011), and other studies have showed that cigarette smoking may play the most harmful role in the initial development of GC, and that drinking alcohol may promote the process (Chen et al., 2000, Sauvaget et al., 2005, Yamaji et al., 2009).So smoking and alcohol should be avoided or diminished in order to prevent GC incidence.For other diet and lifestyle factors, available evidence showed that the relationships between preserved fish, preserved vegetable, smoked food, green tea, red meat, beer, glycemic load, vitamin C, ham, sausage and GC were inconsistent, and there were not significant associations between glycemic index, folate, coffee, fish, omega-3 fatty acids, onion leaf and GC.
However, for some factors, such as omega-3 fatty acids, the number of studies that evaluated their relationships to GC was few and the conclusion could not be drawn.And for inconsistent factors, such as green tea and glycemic load, the different results from several SRs or meta-analyses or primary studies (cohort or CC studies) were found.There are two SRs (Borrelli et al., 2004;Boehm et al., 2009) and three meta-analyses (Zhou et al., 2008;Myung et al., 2009;Kang et al., 2010), which evaluated the association between green tea consumption and GC risk.There were discrepancies in the effects of green tea consumption on GC risk between CC and cohort studies in three meta-analyses (Zhou et al., 2008;Myung et al., 2009;Kang et al., 2010), as well as between the crude and adjusted data in the pooled results (Myung 2009;Kang 2010).So we have to pay attention to several things.Firstly, these three meta-analyses (Zhou et al., 2008;Myung et al., 2009;Kang et al., 2010) included both case control (CC) and cohort studies, but their pooled results of these two kinds of studies were not consistent.In this condition, among the pooled results of CC studies, or cohort studies, or all studies, which one should we believe?We suggested that if there was no heterogeneity between all studies, and across the subgroup analyses of these two study designs, we believed the pooled results of all studies.
If not, we believed the pooled results of cohort studies, as in the evidence pyramid cohort studies were more believable than CC studies, and further cohort studies were needed to confirm the results of meta-analysis.Secondly, the first meta-analysis (Myung et al., 2009), which used both crude and adjusted data, showed differences between these results.However, which result should be trusted?We suggested the pooled results of adjusted data should be more believable if the adjusted data was right, as confounders in included studies were adjusted.But the discrepancies in the effects of green tea consumption on GC risk between CC and cohort studies, or between the crude data and adjusted data make us confused, so further meta-analyses with rigor methodological ways and comprehensive studies included were needed.Also, further cohort studies were needed to confirm the results of these meta-analyses.As a result, more studies (including SRs, meta-analyses, cohort or CC studies) were needed to provide more information.
For co-morbid conditions, overweight or obese could increase GC risk, asthma could decrease GC risk (one study), and the relationship between diabetes (two studies) and GC was inconsistent.So suggestions about avoiding excessive weight gain, achieving and maintaining a healthy weight if currently overweight or obese, were made.But for the other two factors, we could not conclude any conclusions and make any suggestions.
For drug use, available evidence showed that there was not a significant relationship between statins use and GC risk, and the relationships between NSAIDs use (including aspirin) and GC risk were inconsistent.Two metaanalyses (Tian et al., 2010;Yang et al., 2010) evaluated the relationships between aspirin use and GC risk, but they were different in their ORs, as one meta-analysis included one RCT (Cook et al., 2005) and another PCC study (Figueroa et al., 2009).So the other meta analysis (Yang et al., 2010) were much more comprehensive than the first one (Tian et al., 2010).That is the two studies, which might change OR between aspirin and GC.So we are not sure that they may change the conclusion of the first meta-analysis, if the two studies (Cook et al., 2005;Figueroa et al., 2009) mentioned above were included.Why did these happen?The first review (Tian et al., 2010) searched three common medical databases (Medline, Embase, and Web of science) in March 2009, but it did not find the PCC study (Figueroa et al., 2009), which was epub 2008 Nov 7. We searched PubMed for this paper (Figueroa et al., 2009), and found this paper was indexed in PubMed and free to download.We thought it is not included as omitting by the study selectors and that might be the reason why the authors of this paper did not mention the process of study selection.So in the future, more rigor methods and comprehensive searches should be applied in SRs/meta-analyses making.
For occupational factors, available evidence showed that Pb could increase GC risk, and there were not significant associations between occupational exposure to Cr, chrysotile fiber and GC.So we suggested keeping away from occupational exposure to Pb, or taking enough protective actions when you have to.
Regarding methodological qualities of included SRs/ meta-analyses, for 55 included SRs or meta-analyses that evaluated the GC etiology, their methodological qualities of included SRs or meta-analyses were not so good and nearly half of them were low quality.Very few studies conducted all contents of item 2, item 8, item 10, and item 11.And 9.09% to 92.73% studies did not conduct any contents of item 2 to item 11, especially for item 2, 4, 7, 8, which were not conducted in more than half of studies.Of 52 small items which were operationalized from the R-AMSTAR (Kung et al., 2010), seven small items were not conducted in all included studies , twenty-seven small items were conducted in less than half of all included studies (eleven small items were conducted in less than 25% of all included studies).Even though, meta-analyses/ SRs of observational studies about the etiology have been developed for about twenty years, their methodological quality were still low.
One terrible problem that exists in meta-analyses/ SRs of observational studies about the etiology of GC was the search, study selection, and data extract.Nearly half of studies searched only one electronic source, such as Pubmed.Till now, we could not know whether only searching in Pubmed could find all available studies about one topic or not.But we know that, only searching in Pubmed might lose to find some studies that were only indexed in other databases, such as Embase.Also given that articles are often misclassified, this study also highlights the importance of searching various databases in addition to other methods such as snowballing reference list of included studies (Anne).However, 47.26% only searched Pubmed to provide evidence, and 32.73% did not search by reviewing the references in the studies found or consulting current contents, reviews, textbooks, specialized registers, or experts in the particular field.We could not say that this search of one database without tracking reference list must miss some studies in their review, but the kind of search strategy might loss some studies that met their inclusion criteria.56.36% studies did not do duplicate study selection and data extraction, while only 27.27% studies did duplicate study selection and data extraction.This is so terrible.One person or two dependent persons might make some mistakes during data extracting, which is the reason why well conducting SRs/metaanslyses need two independent persons to select studies and abstract data.Even though, high prevalence of data extraction and reporting errors were found in Cochrane systematic reviews (Jones et al., 2005).We compared the search strategy, study selection, data abstraction of three SRs/meta-anslyses about the relationships between green tea and GC risk (Zhou et al., 2008;Myung et al., 2009;Kang et al., 2010).Three meta-analyses (Zhou et al., 2008;Myung et al., 2009;Kang et al., 2010) we reviewed did not include the same studies, even though their results were similar.These three meta-analyses (Zhou et al., 2008;Myung et al., 2009;Kang et al., 2010) searched PubMed and the Cochrane library after 2006, and the inclusion criteria of both meta-analyses were similar, but all failed to include some studies before 2006.We wondered how it happened.One meta-analysis (Myung et al., 2009) reported that all of the studies retrieved from the databases were independently evaluated by 3 evaluators, but the other two (Zhou et al., 2008;Kang et al., 2010) did not mention who selected the studies for eligibility.Each meta-analysis (Zhou et al., 2008;Myung et al., 2009) has one study missing to be found.Although the third metaanalysis (Kang et al., 2010) included both Chinese and Japanese studies, but it failed to include some studies the previous two meta-analyses included (Zhou et al., 2008;Myung et al., 2009).Why did this happen?This might be because they failed to search them or excluded them during selecting.At the same time, we could find some mistakes about data abstracting in one meta-analysis.This also happened in the topic about the relationships between aspirin use and GC risk (Tian et al., 2010;Yang et al., 2010).One meta-analysis (Yang et al., 2010) included one RCT (Cook et al., 2005) and another PCC study (Figueroa et al., 2009) than the first one (Tian et al., 2010).That is the two studies, which changed OR between aspirin and GC.So we are not sure that they may change the conclusion of the first meta-analysis, if the two studies (Cook et al., 2005;Figueroa et al., 2009) mentioned above were included.Why did these happen?The first review (Tian et al., 2010) searched three common medical databases (Medline, Embase, and Web of science) in March 2009, but it did not find the PCC study (Figueroa et al., 2009), which was epub 2008 Nov 7. We searched PubMed for this paper (Figueroa et al., 2009), and found this paper was indexed in PubMed and free to download.We thought it is not included as omitting by the study selectors and that might be the reason why the authors of this paper did not mention the process of study selection.So in the future, more rigor methods and comprehensive searches should be applied during SRs/meta-analyses making.So how to prevent mistakes in finding some studies and data abstraction from happening again deserves thinking and needs actions, such as search more electronic sources and reference tracking, discussing the disagreements among reviewers again and again.We suggested that it is better to search at least electronic sources along with reference tracking, and select studies, extract data by two independent persons in the process of conducting SRs/ meta-analyses of observational studies.
Another terrible problem is that 81.82% did not assess the scientific quality of their included studies, and 92.73% did not use the scientific quality of the included studies in formulating conclusions.We all know, observational studies began with low quality (Balshem et al., 2011).If we evaluated the scientific quality of the included studies without any other downgrading factors, we could be more confident with the results.If the methodological quality of included studies was not evaluated, we could not judge the results could be changed if there were limitations.So the results of assessment could influence the confidence of the conclusions.However, available evidence showed that almost all SRs/meta-analyses of observational studies about the etiology did not evaluate the scientific quality their included studies.Why could it be like that?There may be several reasons: first, more instruments were used to evaluate the methodological quality of observational studies, such as the Newcastle-Ottawa Scale (NOS), or methodological index for non-randomized studies (MINORS) (Wells et al., 2000;Slim et al., 2003;Stang et al., 2010), and it is hard to choose an appropriate one.Second, it is difficult to use them, as the authors did not have the ability to understand the contents or it will add the burden of working.Third, the authors were not aware of evaluating the scientific quality of their included studies or they did not know that there were these instruments.So in the future, we suggested using appropriate instrument to assess the methodological quality and implant the assessment results in formulating conclusions.
Overall, the methodological qualities of included SRs/ meta-analyses were not so good, as some did not report necessary items.So the validities of the results of these SRs/meta-analyses were questionable, as some serious methodological flaws might lead to a high risk of bias (Lundh and Knijnenburg, 2009).In the future, checklists about assessing the methodological qualities and reporting SRs/meta-analyses were needed to follow (Stroup et al., 2000;Shea et al., 2007;Kung et al., 2010 ).
Rearding strengths and limitations, our overview is the first one that comprehensively reviewed all available SRs/meta-analyses about the risk and protective factors of GC using rigor evidence based medicine methods: employ objective searches of the literature, apply predetermined inclusion criteria, critical appraisal of all relevant studies.Also we used GRADE approach to evaluate the quality of evidence and gave recommendations about GC prevention, which was the first paper to give recommendations based on SRs and meta-analyses using evidence based methods and GRADE approach.Even though, our work has potential limitations.Firstly, we just included SRs or meta-analyses, which means we did not review the factors of GC in primary studies (such as cohort studies, CC studies, etc).As a result, we might miss some factors of GC.Secondly, we just reviewed the single factor of GC, so the interactions between two or more factors for GC were not included.Thirdly, the methodological qualities of included SRs/meta-analyses were not so good, as some did not report necessary items, such as data extracting process.So the validities of the results of these SRs/meta-analyses were questionable, and some serious methodological flaws might lead to a high risk of bias (Lundh and Knijnenburg, 2009).Fourthly, there were so few primary studies about one topic that the conclusion about it was less confirmable.The last limitation was the big problem of our study.Only four SRs/ meta-analyses of preventing GC were found (Bjelakovic et al., 2008;Fuccio et al., 2009;Druesne-Pecollo et al., 2010;Shimoyama et al., 2011), but there were 55 SRs/ meta-analyses about etiological factors.Here we must make clear an important distinction between evidence on the GC etiology and evidence supporting the effectiveness of preventive actions.The results from etiological research or addressing the impact of interventions cannot be taken as equivalent in terms of their support to preventive actions.For example, the evidence that H. pylori infection is associated with GC does not necessarily translate into the conclusion that its eradication will prevent GC.The consequences of such an intervention should also be taken into account when making a recommendation.However, we also gave recommendations when there were not any preventive SRs/meta-analyses, as everyone knows that it is not ethical to expose one group people to the GC risk factors in RCTs.
Regarding implications, for practice, although little evidence was provided, we must pay more attention to how to prevent GC cancer from such small clues.The GC inconsistent factors included diabetes, NSAIDs (aspirin), preserved fish, preserved vegetables, smoked DOI:http://dx.doi.org/10.7314/APJCP.2012.13.5.2069 Systematic Review about Gastric Cancer Risk andProtective Factors food, green tea, red meat, beer, glycemic load, vitamin C, ham, sausage, the GC non-significant factors included occupational exposure to Cr and chrysotile fiber, dupA H. pylori positive strains, glycemic index, folate, coffee, fish, selenium or statins, omega-3 fatty acids, vitamin E, aspirin, onion leaf.Beta-carotene below 20 mg/d, fruit, vegetables, allium vegetables, soybean products, tofu, non-fermented soy-foods, whole-grain, or dairy product, and asthma were the protective factors of GC, while betacarotene 20 mg/d or above, pickled vegetables, fermented soy-foods (miso), processed meat 30 g/d or above, or bacon, or salty foods, exposure to alcohol (50 g/day, 75 g/day, 100 g/day, 125 g/day) or smoking, occupational exposure to Pb, overweight and obesity, helicobacter pylori (including cagA-positive strains) infection were GC risk factors.So we suggested screening and treating H. pylori infection, and limiting the amount of intake of processed meat consumption (30 g/d or below), beta-carotene (20 mg/d and below), pickled vegetables, fermented soy-foods (including miso), and salt or salty foods, lowering alcohol consumption, stopping smoking, avoiding excessive weight gain, keeping away from occupational exposure to Pb, and increasing the quantity of consumption of fresh fruit and vegetable (including allium vegetables), non-fermented soy-foods (including soybean products, tofu), whole-grain consumption, dairy product consumption in dietary food could decrease GC risk in daily life.
For research, the methodological qualities of included SRs/meta-analyses were not so good, and some necessary items were not well conducted, so it needs to be improved according some checklists, such as AMSTER.And there were so few primary studies about one topic that made the conclusion less confirmable, as a result more primary studies about the factors related to GC and how to prevent GC should be conducted.Another terrible thing we have discussed above is the conflicted results of different SRs/ meta-analyses on the same topic as some of them were missed during searching and study selecting.That is why future similar researches need more rigorous methods and ways during systematic review making.