Background

Gastric cancer is the fourth most common cause of cancer-related death and the fifth most prevalent malignant cancer worldwide. Although declining rates, it is predicted that the global burden of stomach cancer would increase by 62% by 2040 [1]. It has been established that gastric cancer has a complex aetiology. Numerous studies have identified host-related factors, environmental factors, and Helicobacter pylori (H. pylori) colonization as risk factors for the development of gastric cancer, which modify the nature and extent of gastrointestinal disorders [2, 3]. Only 1–2% of H. pylori-positive patients developed distal gastric cancer, despite the fact that H. pylori infection may be the cause of a variety of gastrointestinal disorders [3]. As such, there may be a wide range of host genetic factors which affect an individual’s susceptibility to H. pylori infection. Studies reported that host genetic and epigenetic alterations could result in oncogenic overactivation and tumor suppressor pathways inactivation, leading to gastric carcinogenesis [4, 5].

Innate immunity serves as the first line of defence for human cells against foreign agents. A general differentiation between self and microbial non-self may be drawn by this innate immunity [6]. As a result, the innate immunity that exists in the stomach mucosa is crucial for the development of adaptive immune responses against H. pylori [7]. A family of pattern recognition receptors with a long evolutionary history is the toll-like receptors (TLRs). They were the first family of proteins, in fact, to exhibit Janeway’s predictions that would characterize pattern recognition receptors (PRRs) [8]. Ten receptors are involved in the detection of specific pathogen-associated molecular patterns (PAMPs). For instance, TLR4 is activated by lipopolysaccharide [9], TLR5 is activated by flagellin [10], and TLR9 is activated by unmethylated CpG patterns of bacterial and viral DNA [11], as previously described [12, 13].

Published Genome-Wide Association Studies (GWASs) of gastric cancer in the HuGE PubLit database and the GWAS CatLog of National Human Genome Research Institute (NHGRI) have not examined TLR 9 and its variants. We are aware of the meta-analysis that assessed TLR 2 and TLR 4 in the risk of gastric cancer [13]. Studies on TLR 9 polymorphisms and their relationship with gastric cancer risk emerged in relation to its potentials for initiating an immune response in the presence of H. pylori [14]. The results from these studies are inconsistent. Individual studies that assessed the link between TLR 9 and gastric cancer risk featured a variation in sample size, geographical locations, and ethnic groups, among others. Meta-analysis is a quantitative approach for combining results from various studies on the same topic, and for estimating and explaining their diversity [15, 16]. It is growing as a popular method for resolving discrepancies in genetic association studies [17]. Meta-analysis of genetic association studies is regarded as decisive evidence when carried out properly [18]. We are aware of a meta-analysis with TLR 9 and the cancer risks, where only two individual case-control studies with gastric cancer were included [19]. More individual case-control studies of TLR 9 (-1237 T/C, -1486 T/C) polymorphisms in the gastric cancer risk were available after releasing the Zhang review [19]. Taken together, the objective of this meta-analysis was to synthesize the evidence on the association between polymorphisms of TLR 9 (-1237 T/C, -1486 T/C) and the risk of gastric cancer using data from eligible studies. The finding will help decision makers to identify a high risk group for screening for gastric cancer.

Methods

We followed PRISMA 2020 standards for the reporting of this review (Additional file 1). This meta-analysis study was approved by the Ethics Review Committee of the International Medical University (IMU) in Malaysia (ID: IMU I/BMS1/2021(02). A protocol of the current meta-analysis is available from the corresponding author on reasonable request. This study solely used published data, and therefore the need for consent from participants was waived by the Institutional Ethics Review Committee.

Study selection

We searched relevant studies in electronic databases of Medline via the PubMed interface, EMBASE, EBSCOHOST, Science Direct, and Google scholar. The search terms were “toll-like receptor 9”, “toll 9 receptor” “TLR9” “TLR 9 rs5743836”, “TLR 9 rs187084, “TLR 9” “TLR 9 -1237”, “TLR 9 -1486”, “gastric cancer”, with an appropriate Boolen operator (AND/ OR). The last date of search was December 2022. We also applied the snowball method using manual cross-referencing from retrieved articles to ensure a comprehensive search. Single nucleotide polymorphism identification numbers (SNPs) (rs numbers) were also applied for identification of the eligible studies. Details of the MeSH and terms used are presented in Additional file 2.

Selection criteria

Individual studies were selected based on the following criteria: (1) human study of case- control design, (2) inclusion of cases with confirmed gastric cancer, and controls of gastric cancer -free participants, (3) sufficient information on frequencies of SNPs in both cases and controls − 1237 T/C (rs5743836, rs574383), -1468 T/C (rs352139, rs187084, rs41308230, rs5743844), and (4) use of DNA-based method for genotyping.

To be eligible, a study must provide sufficient information to extract (calculate) an odds ratio (OR) and its 95% confidence intervals (CI). Studies were not considered if they did not meet the inclusion criteria. Hence, non-empirical studies such as editorials, letters to editors, and methodological studies were not considered. Studies of descriptive case series, clinical interventions, preclinical studies or epidemiologic studies such as risk factor analysis of gastric cancer were also excluded.

Data extraction

Two investigators (YZQ and HHA) independently selected the included studies through a four-phase selection process as displayed in the PRISMA-2020 flowchart. The two investigators independently collected data from each study with a use of piloted data extraction sheets. Collected data were: first author, year of study, country of study, tumour location (cardia, non-cardia), ethnicity, SNP types, detection methods, the total number of cases and controls, frequencies of alleles and genotypic distributions in cases and controls, status of Hardy-Weinberg equilibrium (HWE), and minor allele frequency (MAF). If MAF was not provided in the study, we derived it. If HWE was not provided in the study, it was assessed in the controls using a goodness of fit chi-square test; a p value less than 0.05 was deemed significant disequilibrium [20]. Any disagreements in these steps were resolved by discussion with the third investigator (SNA).

Quality assessment

The methodological quality of the studies was evaluated with the Newcastle-Ottawa Scale (NOS) criteria [21]. This tool evaluates three main domains: the selection and representativeness of cases and controls, comparability of cases and controls, and the methods of ascertainment for cases and controls. The NOS criteria’s quality score ranges from zero to nine, in which a higher score represents a better-quality study. Two reviewers (YZQ and HHA) assessed a quality score for each item. Any disagreement between the two investigators was resolved by discussion with the third investigator (SNA).

Statistical analysis

For each study identified, study-specific variance of the natural logarithm of adjusted OR and its 95% CI were derived to measure the risk effect of gastric cancer and TLR 9 (-1237 T/ C, -1486 T/ C). For the overall pooled analysis (i.e., the main analysis of this meta-analysis), we estimated the summary OR and its 95%CI of all included studies stratified by ethnic groups, regardless of their HWE status. Summary estimates were obtained using the random-effects models (i.e. DerSimonian and Laird), reflecting substantial between-study heterogeneity [22]. Otherwise, a fixed effect model would be used. Between-studies heterogeneity was assessed with Cochran’s Chi2 -based Q statistic (P value for Q statistic < 0.10 is considered significant heterogeneity) [23], and the I2 statistics [24]. The I2 represents the percentage (%) of the observed between-study variability due to heterogeneity rather than to chance. It ranges between 0% and 100%; values above 75% imply substantial heterogeneity [24]. Meta-analysis was performed with five genetic models such as allelic contrast, dominant, recessive, homozygous, and heterozygous models to determine the association between TLR 9 (-1486 T/C, -1237 T/C) and gastric cancer risk. Analysis was stratified by two broad ethnicities such as Asian and non-Asians. The category ‘Asian’ includes all individuals who identify with one or more nationalities or ethnic groups originating in the Far East, Southeast Asia, or the Indian subcontinent [25]. It reflects a geographic location-based category, rather than biological functioning categories. Non-Asians in this study cover those participants other than Asian countries. For robustness of estimates, sensitivity analyses were done with a leave-one out meta-analysis.

A minimum of ten studies is required to conduct Egger’s test for publication bias [ 25]. Publication bias was not done due to the presence of eight studies only. Data analysis was done with RevMan 5.4 (Copenhagen), and meta package of R (version 4.3).

Results

Figure 1 presents the study selection process. The initial search yielded 382 studies from the databases and one additional study via manual searching. Ten duplicate studies were then removed. After screening the titles and abstracts, 346 studies were further excluded. Eighteen full-text studies were checked, and ten studies were excluded for different reasons. A final of eight studies that assessed the association of TLR 9 polymorphisms (-1486 T/C, -1237 T/C) were eligible for this meta-analysis [5, 14, 26,27,28,29,30,31]. The excluded ten studies and their justifications were provided in Additional file 3.

Fig. 1
figure 1

Study selection process

Full size image

Study characteristics and quality assessment

Eight studies (nine datasets), incorporating 3,644 participants (1,914 cases, 1,730 controls) across nine countries that assessed the association of two TLR 9 polymorphisms (-1486 T/C, -1237 T/C) were eligible for the current analysis. The main characteristics of the eight included studies is provided in Table 1. One study (two datasets) assessed TLR 9 (-1237 T/C) [14], five studies assessed TLR 9 (-1486 T/C) [5, 26, 29,30,31], and two studies assessed both SNPs (-1237 T/C and − 1496 T/C) [27, 28]. Half of the studies were conducted in a single country in the Asian region such as China [5, 26, 29, 30], while four studies in the non-Asian countries such as Brazil [27], Egypt [31], Italy [28], and Poland as well as the USA [14]. These eight studies were published between 2009 and 2022. Seven studies were hospital-based case-control designs [5, 26,27,28,29,30,31], and only one was a population-based study [14]. Half of these studies used PCR-RFLP [5, 26, 30, 31], and another half used TaqMan for genotyping [14, 27,28,29]. For TLR 9 (-1237 T/C), only one study [27] violated HWE in the control participants. This happened in two studies of TLR 9 (-1486 T/C) [5, 31]. The NOS methodological quality assessment rated the included studies ranged from six to eight stars (Additional file 4), in which only three studies were of high quality (i.e., ≥ 7) [26, 29, 31]. The genotype frequencies of each study identified in the current meta-analysis are provided in Table 2. The distribution of MAF among the studies ranged from 9 to 42.5%, reflecting that none was ‘monomorphic’ [32] and all were ‘common’ distribution.

Table 1 Characteristics of the included studies
Full size table
Table 2 Distribution of genotype frequencies of included studies
Full size table

TLR 9 (-1486 T/C) and the risk of gastric cancer

Overall, six studies (i.e., Asian ethnic group: 4 studies, non-Asian ethnic group: 2 studies) investigated the association between polymorphism of TLR 9 (-1486 T/C) and the gastric cancer risk [5, 26, 27, 29,30,31]. All except one study [33] satisfied the HWE. Under the dominant model, it was marginally and significantly associated with the risk of gastric cancer only in the Asian ethnic subgroup (overall: OR = 1.22, 95%CI = 0.90–1.67, I2 = 56%; Asian: OR = 1.24, 95%CI = 1.00–1.54, I2 = 0%, non-Asian: OR = 1.25, 95%CI = 0.38–4.09, I2 = 89%) (Fig. 2). Under the recessive model, there was a significantly increased risk of gastric cancer only in the Asian group (overall: OR = 1.4, 95%C I = 0.74–2.64, I2 = 85%; Asian: OR: 1.41, 95% CI = 1.07–1.86, I2 = 0%, non-Asian: OR = 1.18, 95%CI = 0.12–11.76, I2 = 97%) (Additional file 5). Under the heterozygous model, there was no significant association with the risk of gastric cancer in overall or any ethnic subgroup (overall: OR = 1.12, 95% CI = 0.92–1.36, I2 = 0%; Asian: OR = 1.16, 95% CI = 0.92–1.45, I2 = 0%; non-Asian: OR = 0.99, 95% CI = 0.65–1.51, I2 = 0%). Under the homozygous model, a significantly increased the risk of gastric cancer was observed only in the Asian group (overall, OR = 1.47, 95% CI = 0.76–2.86, I2 = 82%; Asian: OR = 1.54, 95% CI = 1.13–2.1, I2 = 0%; non-Asian: OR = 1.19, 95% CI = 0.1-14.33, I2 = 96%) (Additional file 5). Under the Allele model, a significantly increased the risk of gastric cancer was observed only in the Asian group (overall: OR = 1.23, 95% CI = 0.89–1.71, I2 = 84%; Asian: OR = 1.22, 95% CI = 1.05–1.41, I2 = 0%; non-Asian: OR = 1.24, 95% CI = 0.34–4.59, I2 = 97%) (Additional file 5).

Fig. 2
figure 2

Forest plot showing the effect estimates for TLR 9-1486 (T/C)

Full size image

TLR 9 (-1237 T/C) and the risk of gastric cancer

Overall, four studies with five datasets (i.e., Asian ethnic group: 1 study, non-Asian ethnic group: 3 studies) assessed the association between this polymorphism and the gastric cancer risk [14, 27, 28, 30]. Under the dominant model, there was no significant association between TLR 9 (-1237 T/C) and the risk of gastric cancer in overall or any ethnic subgroups (overall: OR = 1.39, 95%CI = 0.78–2.48, I2= 85%; Asian: OR = 3.19, 95%CI = 0.96–10.56, I2 = 0%, non-Asian: OR = 1.23, 95%CI = 0.67–2.28, I2 = 87%) (Fig. 3). So were the remaining four genetic models (Additional file 6) and (Table 3).

Fig. 3
figure 3

Forest plot showing the effect estimates for TLR 9-1237 (T/C)

Full size image
Table 3 The effect estimates for TLR 9-1486 T/C and − 1237 T/C
Full size table

Sensitivity analyses

We conducted a leave-one-out sensitivity analysis by sequentially discarding one study at a time due to variations in sample size, genotyping techniques, and the presence of small studies. The results of the sensitivity analysis revealed that the effect estimates were unstable. For instance, TLR 9 (-1237 T/C) and the risk of gastric cancer among non-Asians and the overall group were significantly related under the allele model (i.e., after omitting a study [30]). None had been determined to be significantly associated in the initial analysis. This demonstrated that the estimates were unstable. In the recessive, homozygous, and allele models (i.e. after omission), TLR 9 (-1486 T/C) and the risk of gastric cancer was significantly related in Asians, non-Asians, and the whole population. This was different from the findings in the initial analyses (Table 4), reflecting the instability of the estimates.

The estimates for all five genetic models were reanalysed after a study that departed from HWE in TL9 (-1486T/C) [31] was removed, and the different findings were shown. As an example, under the recessive model, significant associations have been observed in both Asian and non-Asian groups with different directions (overall, OR = 1.06; 95%CI = 0.83–1.35; Asian: OR: 1.42; 95%CI = 1.08–1.47; non-Asian: OR = 0.37; 95%CI = 0.21–0.64). The same was true for the other four genetic models (Table 4). This implied HWE deviation also had an impact on the effect estimates.

Table 4 The estimates of a leave-one- out meta-analysis
Full size table

Discussion

Given the association between inflammation and carcinogenesis, candidate gene approaches are becoming more and more attractive for identifying genes that may initiate and progress inflammation-associated carcinogenesis, particularly gastric cancer, in the gastrointestinal tract. In this meta-analysis, we evaluated the influence of two prominent SNPs (1237 C/T and − 1486 T/C) in the TLR9 gene on the risk of gastric cancer across six countries.

A previous review that included two case-control studies on gastric cancer revealed a lack of statistical significance between TLR 9 (-1237 T/C) and the risk of gastric cancer employing the selected inheritance model [15]. This is comparable to our analysis of the four studies that satisfied the HWE. This means that independent of the included studies and total samples, TLR 9 (-1237 T/C) has no functional role in the development of gastric cancer. In our study, there were no reliable estimates of the association with TLR9 (-1486 T/C). In general, the results that are not significant may be attributed to (1) the inability to adjust the conventional risk factors, such as a family history of gastric cancer, or common risk factor such as H. Pylori infection [2] a lack of power to detect a significant association as a consequence of a paucity of studies, and [3] the within-population heterogeneity, geographic variation, and difference in source population (i.e., hospital-based vs. population-based). It also seems that these SNPs may be more closely related to non-significant gastric cancer risk. There is also a likelihood that these SNPs may have significance for defining the host immune response to H. pylori infection, but it does not appear that they determine what happens subsequently in the development of cancer [14]. Notable are only one Asian study and three studies with Caucasian groups for TLR 9 (-1237 T/C), which might be confounded by population stratification.

Study limitations

There are several limitations that should be acknowledged. First, the sample sizes in the current study were small. For example, less than ten studies were identified, and only one study included the Asian-subgroup in TLR 9 (-1237 T/C). Hence, a type II error is a concern. We acknowledge the issue that there may be more specific ethnic differences that exist amongst populations in the same country. Also, there might be a common confounding factor such as age, gender, H. pylori infection status. Due to the inconsistent manner of reporting, we were unable to do meta regression with common covariates. Moreover, in small studies identified by the current analysis, a statistically significant finding would actually be a false-positive report probability (FRPP). The FRPP, which is the probability of no association between a genetic variant and a disease (i.e., gastric cancer in this case) gives statistically significant results in terms of the observed p value, the prior probability that the association between the genetic variant and the disease (i.e., gastric cancer in this case) is real, and the statistical power of the test [32]. We included only published studies in English. Hence, there might be relevant studies in other languages or non-published studies, which could contribute to ‘information bias’. It was not possible to assess reporting bias by creating a funnel plot because there were fewer than ten trials included. The subgroups of Asians and non-Asians were location-based categories, not solely reflecting biological categories or ethnic differences. This broad classification is an ease of the current analysis, but care must be taken to interpret that the findings did not represent the actual ethnic differences in genotype distribution.

The included studies used different genotyping methods, which might be associated with different genotyping success rates and data quality. However, genotyping errors are expected to be small, and thus the resulting biases are likely to be small [33]. As the aetiology of gastric cancer is complex, the influence of polymorphisms in the TLR 9 gene might be covered by unidentified genes and environmental factors. Hence, the comprehensive analysis of gene-gene interaction and gene-environment interaction might be more informative. It is, however, beyond the scope of our analysis.

Implications for clinical practice

Based on limited data presented in this review, we still do not confirm whether TL9 gene polymorphisms (-1237 T/C and − 1486 T/C) increase or reduce the risk of gastric cancer. Our analysis highlights the need for additional studies on the role of the TLR 9 gene polymorphisms in the risk of gastric cancer, and the findings should be used to inform healthcare providers considering screening for genetic risk factors. Such research could aid in identifying patients who have considerably higher risks of disease progression and may guide the development of customized prevention and management plans for H. Pylori infection.

Conclusion

The findings suggested that TLR9 (-1237 T/C and − 1486 T/C) have some roles in the associated risk of gastric cancer. The information size was inadequate to achieve confirmatory evidence. Future well-designed case-control studies with an adequate number of participants in multi-ethnic groups, stratified with common factors, are recommended to substantiate the evidence on the relationship between these two polymorphisms and gastric cancer risk.