Background

Gastric cancer (GC) is one of the most prevailing cancers and the top three cancer-related death causes worldwide. In 2018, it was estimated that over 782, 000 patients died of gastric cancer, and more than half of the deaths occurred in Eastern Asia [1]. Despite the progress at diagnosis and therapies in recent years, the prognosis of GC is still limited. More studies were warranted to recognize patients at risk of recurrence or death from GC by exploring more biomarkers.

Helicobacter pylori (H. pylori), a bacterium colonizing in the stomach, was graded as a Group I carcinogen in 1994 by the International Agency for Research on Cancer [2]. Extensive studies have concluded that H. pylori infection contributed to gastric cancer, and it is estimated that nearly two-thirds of new gastric cancer cases are attributable to chronic H. pylori infection [3] and the eradication of H. pylori could reduce the risk of GC [4].

Recently, epidemiological studies have suggested that H. pylori infection was also related to the prognosis of GC [5,6,7,8,9,10]. Some studies have indicated that patients positive for H. pylori have better overall survival (OS) compared to patients of negative [5,6,7]. Other studies reported that H. pylori status was not associated with the survival of GC [8,9,10]. There are also studies showing that H. pylori infection has an association with worse overall survival [11,12,13]. Therefore, the relationship between H. pylori infection and the prognosis with respect to GC is still unclear. Previous meta-analyses have examined this issue [14,15,16]. The most recent study by Fang et al. retrieved the relevant studies that were published before March 2017 and showed that H. pylori-positive status was related to longer OS [16]. Recently, more studies on GC prognosis report the status of H. pylori infection, with a better understanding of H. pylori and the availability of detection methods. However, many prognostic studies did not use the words H. pylori in the title, abstract, or keywords sections of the paper [17,18,19], thus limiting the ability to retrieve these studies in a literature search. This phenomenon is more prominent in studies showing H. pylori status is not associated with GC prognosis. Therefore, many studies may be missed when the term H. pylori is used as part of the search strategy, which was done by all the previous meta-analyses. Moreover, 3 years have passed since the last meta-analysis, and ever since, dozens of studies on this issue have been published. Therefore, to comprehensively search for the relevant studies and summarize the relationship between the H. pylori status and GC survival, we performed this study, in which we combined the results of our primary prospective cohort study with the results from a meta-analysis.

Methods

Prospective cohort study

Subjects

Gastric cancer patients were recruited from the First Hospital of Jilin University from January 2009 to December 2013. These patients were histologically diagnosed with gastric cancer and hospitalized for curative-intent tumorectomy, without any chemotherapy before surgery. Their demographical and clinicopathological data were collected. Pathological parameters were determined based on postoperative pathologic examination. The histological type was assessed by the criteria of the World Health Organization and categorized as tubular adenocarcinoma, signet ring cell, and others. Histological grade was defined as well-to-moderate differentiated and poorly differentiated. Clinical stages were determined according to the American Joint Committee on Cancer (7th edition) [20]. All the participants signed the informed consent before entering the study.

Patients were prospectively followed-up periodically after being discharged from the hospital. The follow-ups were scheduled for 3 months, 6 months, 1 year after surgery, and annually afterwards. And information on post-surgery chemotherapy, survival status, including death date and death reason if the patients died, were collected. Survival time was defined as the duration from the date of tumorectomy to the date of death. And when the patients were alive or lost to follow-up, the calculation of the period of survival was based on the date of tumorectomy and the date of the latest successful contact. Patients who died within 1 month of surgery were excluded from further survival analysis.

Detection of H. pylori status

Presurgery levels of serum immunoglobulin G (IgG) antibodies to H. pylori were evaluated by enzyme-linked immunosorbent assay (Biohit, Helsinki, Finland). Titers > 30 EIU were counted as positive for H. pylori according to the kit instructions.

Statistical analysis

Continuous variables like age were described as median with interquartile range and compared by Mann-Whitney U test between H. pylori-positive and -negative groups. Categorical variables were presented as frequencies with proportions and compared with Pearson χ2 test or Fisher’s exact test. The survival curves were plotted using Kaplan-Meier method and compared by the log-rank test between patients of H. pylori-positive and -negative. Multivariate Cox’s proportional hazard regression was utilized to calculate the independent predictive value of H. pylori on GC overall survival after adjusting for other potential prognostic factors by estimating the hazard ratios (HRs) with their 95% confidence intervals (CIs). All analyses were performed using SAS software (version 9.4, SAS Institute, NC, USA). A two-tailed P-value less than 0.05 was considered statistically significant.

Meta-analysis

Search strategy

A comprehensive search was performed in PubMed and Embase databases for studies concerning patients’ prognosis on gastric cancer published in English with the strategy of (1) “stomach” or “gastric”; and (2) “cancer” or “neoplasm” or “carcinoma” and (3) “survival” or “prognosis”. Then, the full-texts of the retrieved articles were assessed to screen all the potential studies on H. pylori and GC prognosis. The search was last conducted on April 10, 2020. Detailed retrieving strategies were attached in the supplementary file 1.

Eligibility criteria of the studies

Studies were eligible if: (1) they evaluated the association of H. pylori status at diagnosis with a prognosis of gastric cancer; (2) the prognosis was about overall survival (the primary outcome of our interest) or recurrence-free survival (RFS) and disease-free survival (DFS) (the secondary outcomes of interest); (3) the studies designed as cohort studies, irrespective of prospective or retrospective cohort; (4) they reported the HRs and 95% CIs to quantify the association, or there was sufficient information to estimate the relevant HRs and 95% CIs. Studies were excluded if they had overlapping subjects with studies already included.

Quality assessment

The quality of each eligible study was evaluated with the modified Newcastle-Ottawa Scale (sTable 1) based on three aspects: selection of subject, comparability between groups, and determination of the outcome. The maximum score was 9, and studies with scores greater than 7 were designated as high-quality. The evaluation was independently performed by two investigators (ZM and ZFJ), and disagreements were resolved by re-evaluating and then discussing with the third investigator (JJ).

Data extraction

The following data was extracted from each eligible study: the first author, publication year, geographical location of subjects, demographic characteristics of subjects, clinical stage of the patients, the method to determine H. pylori, number of patients positive and negative for H. pylori, surgery treatment, duration of follow-up and prognostic evaluation of univariate or multivariate analysis. Two investigators independently extracted all the data (ZM and ZFJ) and disagreements were resolved by group discussion.

Statistical analysis

HRs and their 95% CIs were extracted to quantify the prognostic value of H. pylori infection on gastric cancer. HRs from multivariate Cox regression analysis were preferred. If unavailable, HRs of a univariate analysis would be extracted. If no HRs were available, HRs with 95% CIs would be estimated using the tool developed by Tierney et al. [21] or estimated from Kaplan-Meier survival curves using Engauge Digitizer(version 4.1, http://digitizer.sourceforge.net). Then HRs were pooled using the random-effect model allowing for heterogeneity among studies. Heterogeneity was evaluated by I2 and the P-value of the Q test. If I2 is over 50%, the heterogeneity is thought to be high. If I2 is below 25%, the heterogeneity is thought to be low. Otherwise, the heterogeneity is thought to be moderate. Funnel plots and Egger’s test were applied to examine the potential publication bias of the studies included. Sensitivity analyses were performed by several methods. Galbraith’s plot and leave-one-out analysis were used to display the sources of heterogeneity. HRs were also summarized after excluding studies showing high heterogeneity to test the stability of the pooled results. Subgroup analyses for the relationship of H. pylori infection status with overall survival were conducted according to the study-level factors such as study location, retrieval method, and H. pylori determination method. Cumulative meta-analysis displayed by a forest plot was used to show the change of the pooled results over time. All analyses were conducted in Stata software (version 12.0, Stata Corp, TX, USA), and a two-tailed P-value less than 0.05 was deemed statistically significant.

Results

Cohort study

A total of 743 eligible patients were included in our cohort study (sFigure 1). Among these 743 patients, 516 patients (69.4%) were positive for H. pylori, while 227 (30.6%) were negative. The H. pylori-positive patients had shorter tumor diameters (median 4.0 vs 4.5 cm, P = 0.012), and therefore, patients with larger tumor diameters (> 4.5 cm) had a lower proportion of H. pylori-positive patients (P = 0.042, Table 1). We did not observe any statistically significant association between H. pylori infection status and demographics characteristics, such as age (P = 0.191) and gender (P = 0.088), or other pathological factors, such as depth of invasion (T stage, P = 0.828), metastasis of lymph nodes (N stage, P = 0.549), TNM stage (P = 0.089), or the post-operational chemotherapy rate (P = 0.697).

Table 1 Comparison of H. pylori-positive patients and H. pylori-negative patients
Full size table

After a median following-up time of 7.93 years (95% CI: 7.74–8.02), 369 patients (49.7%) died, 338 (45.5%) were alive, and 36 patients (4.8%) lost to follow-up. The median survival time was 7.43 years (95% CI: 5.78–9.48) and the five-year survival rate was 55.1% (95% CI: 51.4–58.6%) for these 743 patients with curative tumorectomy.

Among 516 patients positive for H. pylori, 250 patients (48.4%) died and the median survival time was 7.68 years. Meanwhile, 119 of the 227 H. pylori-negative patients (52.4%) died and the median survival time was 5.82 years. The survival curve showed a tendency that the H. pylori-positive patients had a slightly lower risk of death, though observed no statistical significance (HR 0.89, 95% CI: 0.72–1.11, P = 0.294, Fig. 1). The same nonsignificant trend was observed on the 5-year overall survival: 56.7% (95% CI:52.3–60.9%) for patients of H. pylori positive while 51.4% (95% CI: 44.6–57.8%) for patients negative (P = 0.185).

Fig. 1
figure 1

Survival plots for gastric cancer patients of H. pylori-positive and -negative

Full size image

To clarify the consistency of the relationship between H. pylori and GC survival, we performed subgroup analysis stratified by potential prognostic factors (Fig. 2). While there is no significant association was observed in any subgroup. Notably, the point estimation showed a slightly larger trend in patients with less tumor length (≤4.5 cm), and earlier clinical stage (T1-T2 stage, N0 stage, and TNM I stage, Fig. 2).

Fig. 2
figure 2

Stratified analysis on the association of H. pylori status with gastric cancer overall survival

Full size image

Further multivariate Cox regression analysis indicated that H. pylori status was not significantly associated with the overall survival of GC patients after adjusting for other prognostic factors (HR 0.92, 95%CI: 0.74–1.15, P = 0.486). Seven variables were independent predictive factors for OS of gastric cancer: older age (> 65 vs ≤65 years, HR 1.48, 95% CI:1.19–1.84, P < 0.001), tumor diameter (> 4.5 vs ≤4.5 cm, HR 1.38, 95%CI: 1.12–1.71, P = 0.003), T stage (T3-T4 vs T1-T2, HR 2.46, 95%CI: 1.72–3.51, P < 0.001); N stage (N1-N3 vs N0, HR: 2.17, 95%CI: 1.55–3.03, P < 0.001), lymphovascular invasion (positive vs negative, HR: 1.82, 95% CI: 1.31–2.55, P < 0.001), neural invasion (positive vs negative, HR: 1.35, 95% CI: 1.06–1.71, P = 0.016), and postoperational chemotherapy (yes vs no, HR 0.77, 95% CI: 0.61–0.96, P = 0.019) (sTable 2).

Meta-analysis

Studies included in the study

The flow chart of the potential studies appears in Fig. 3. The search of PubMed and Embase databases yielded 14,661 studies. After screening the full texts of these studies, 165 studies reported results on the association of H. pylori with GC prognosis. Among them, 59 studies fully met the inclusion and exclusion criteria and were included in the meta-analysis (Table 2).

Fig. 3
figure 3

Flow chart of the screening procedure for studies included in the meta-analysis. No terms on “H. pylori”: studies without relative terms of “H. pylori”; No sufficient: no sufficient data to transform into hazards ratio estimate. Other: studies were excluded by one of the reasons below: (1) some of the subjects were gastric lymphoma; (2) data were from the public databases like TCGA (The Cancer Genome Atlas); (3) the study did not report the relationship of H. pylori with a prognosis of gastric cancer

Full size image
Table 2 Baseline characteristics of the 59 studies included in the meta-analysis
Full size table

These 59 published studies [5, 6, 8,9,10,11,12,13, 17,18,19, 22,23,24,25,26,27,28,29,30, 32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70], together with our primary study, contained 18,315 GC patients (sFigure 2), with a median sample size of 216 (range: 52–1412). The median percentage of H. pylori-positive patients in these studies was 55.4%, ranging from 8.96 to 91.3%. The maximum following-up duration ranged from 2.50 years to 18.33 years. Among these 59 published studies, 18 studies (30.5%) could not be retrieved with the search term H. pylori. For the outcome measurements, 56 studies examined overall survival [5, 6, 8,9,10,11,12,13, 17,18,19, 22,23,24,25,26,27,28,29,30, 32,33,34,35,36,37, 39,40,41,42,43,44,45,46, 48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70], 10 studies examined the disease-free survival (DFS) [9, 24, 27, 34, 37, 39, 48, 57, 61, 63] and 5 studies examined the recurrence-free survival (RFS) [11, 38, 47, 67, 70]. Quality assessment showed that 33 studies had NOS scores > 7 and were classed as a high-quality group, and 26 studies with NOS scores ≤7 were classified as the low-quality group. For the locations of the subjects, 45 studies with 14,164 patients were performed in East Asia, and 14 studies with 3408 patients were carried out in other regions. For the H. pylori determination methods, 20 studies used histological-based methods such as hematoxylin-eosin staining, immunohistochemistry, PCR, or urea breath test; 7 studies used the serum-based methods such as ELISA; 6 studies used methods based on more than one kind of specimen; 16 studies did not report the exact method of detection.

H. pylori-positive status predicted better overall survival

A total of 57 studies, including 56 previously published and our primary cohort study, were included in the meta-analysis on the overall survival of GC. These 57 studies included 17,728 gastric cancer patients. The random-effect model was utilized to summarize the results accounting for the heterogeneity among these studies (I2 = 73.8%, P < 0.001). The pooled HR was 0.81 (95% CI: 0.72–0.90, P < 0.001), indicating that GC patients of H. pylori-positive had better OS than patients of negative (Fig. 4). There was no significant bias as shown in funnel plots and Egger’s test (P = 0.734, Fig. 5).

Fig. 4
figure 4

Forest plot on the association of H. pylori with OS for gastric cancer patients

Full size image
Fig. 5
figure 5

Funnel plot of evaluation of study bias on the association of H. pylori status at diagnosis with OS (A) and DFS (B) for gastric cancer patients

Full size image

Given the significant heterogeneity among studies (I2 = 72.8%; Q = 205.85, P < 0.001), we used a graphical method, i.e., Galbraith’s plot, to identify the possible source of the heterogeneity. The Galbraith’s plot suggested that 6 studies [11, 12, 44, 58, 64, 68] were significant sources of heterogeneity (sFigure 3). When these 6 studies were excluded, we observed a significant reduction in heterogeneity (I2 = 43.3%; Q = 88.18, P = 0.001), while the pooled effect did not change substantially (HR 0.80, 95%CI: 0.73–0.88, P < 0.001) (sFigure 6). The leave-one-out analysis also did not show a substantial change of the pooled effect (sFigure 4), and cumulative meta-analysis also indicated that the summarized effect was temporally stable (sFigure 5).

Further stratified analyses were conducted to examine the stability of the overall estimate and identify the potential study-level influential factors. When stratified by retrieval method, the pooled effect from 17 pieces of research (HR 0.86, 95%CI: 0.73–1.00), which could not be retrieved using the relevant terms related to H. pylori and could only be obtained by checking the full texts, was weaker than the effect among studies retrieved by using the terms of H. pylori (HR 0.78, 95% CI: 0.66–0.91. Table 3). The results from the subgroup of the 25 low-quality studies (NOS scores≤7), of which the lower scores were mainly caused by not reporting the H. pylori determination method and the HRs with 95% CIs, showed that H. pylori infection was not significantly associated with the OS of GC (HR 0.90, 95% CI: 0.79–1.01). When stratified by the method of H. pylori determination, a similar result was observed in the subgroup that did not describe the exact method of H. pylori detection (HR 0.91, 95% CI: 0.77–1.04), and further meta-regression analysis also supported this observation. The results stratified by other factors, such as the geographic location of the subjects (East Asia vs others), the sample size of the study (< 200 vs ≥200), source of HRs with 95% CIs (univariate analysis vs multivariate analysis), H. pylori-positive rate (< 70% vs ≥70%), and surgery treatment (all patients having curative resection vs others), showed that the pooled effects were relatively stable across these subgroups, although the absolute point estimation varied (Table 3).

Table 3 Subgroup analyses for the association of H. pylori status at diagnosis with OS of 56 included studies in the meta-analysis
Full size table

Positive H. pylori status predicted better disease-free survival

We combined the 10 studies on disease-free survival (DFS, 2221 patients) and 5 studies on recurrence-free survival (RFS, 890 patients) into DFS. The summarized results suggested that the H. pylori-positive status might be related to a better short-term outcome of DFS (HR 0.83, 95% CI: 0.67–0.99, Fig. 6), compared to a negative H. pylori status. Funnel plots and Egger’s tests revealed that these studies did not indicate significant publication bias (P = 0.960, Fig. 5).

Fig. 6
figure 6

Forest plot for the association of H. pylori status at diagnosis with DFS for GC patients

Full size image

Discussion

H. pylori infection can result in chronic gastritis, which can progress to gastric atrophy, intestinal metaplasia, and dysplasia, and subsequently to gastric cancer. Our recent research also observed a significant higher H. pylori-positive rate of GC patients than that of nontumor controls. Previous studies on GC suggest a correlation between H. pylori status and the prognosis of gastric cancer, but there are no concrete conclusions regarding this association. In this study, by combining the results from our primary cohort study and pooled estimates from a meta-analysis, we reported that positive H. pylori status associated with a better prognosis of gastric cancer and that the risk of death was 18% lower among H. pylori-positive patients than H. pylori-negative ones (HR 0.82, 95% CI:0.72–0.91).

In our primary cohort study, we observed the trend that the survival of H. pylori-positive patients was better than that of H. pylori-negative patients from the survival plot (Fig. 1). However, the difference did not reach the significant threshold. Though the relatively large number of patients included (N = 743), the post hoc analysis of power pointed out that the test efficacy was only 23.7%, implying an inadequate power at the interpretation of the effect size of HR = 0.92 at the current sample size for our study. To increase the test power by obtaining a much larger sample size, we combined the results of the relevant researches to reach a convincing conclusion.

We used an exhaustive strategy to search for potential studies by checking all the full papers on the prognosis of gastric cancer. This allowed us to find 18 studies not referring to H. pylori in the abstract that were not included in the previous meta-analyses [15, 16]. Three of these papers examined DFS [34, 38, 39] and 17 examined OS [8, 17,18,19, 23, 26, 29, 30, 32,33,34, 39, 40, 44, 45, 49, 60]. Although subgroup analysis showed that these papers did not change the conclusion substantially, the association magnitude was relatively lower (Table 3). Therefore, searching strategy is critically important to cover all the relevant studies in systematic reviews.

In the pooled analysis, we observed that H. pylori-positive patients had longer OS and DFS than H. pylori-negative patients. We also observed the protective role of H. pylori on OS both in East Asian and non-East Asian patients, which differed from the findings of the recent meta-analyses that included a relatively small number of studies [15, 16]. For the detection method, however, we observed that the subgroup of studies that did not describe the exact method of detection indicated no association between H. pylori status and prognosis of GC patients, while subgroups of studies using serum-based or tissue-based methods of detection showed very similar associations (Table 3).

The mechanisms underlying the association between positive H. pylori status and better prognosis have yet to be elucidated. Two explanations may partially explain the association. The first is that H. pylori-related GC might be different from that of non- H. pylori-related GC [71, 72]. The expression profiles of H. pylori-positive GC are different from those of H. pylori-negative GC at both the mRNA and protein levels. These genes are involved in cancer-related pathways such as the ERK/MAPK, and Wnt/β-catenin signaling pathways, which may result from the epigenetic changes and the chronic inflammation induced by H. pylori infection. The second is the immunologic response induced by H. pylori. Bacterial virulence factors of H. pylori, such as cytotoxin-associated gene A (CagA) and vacuolating cytotoxin A (VacA), not only facilitate colonization of the bacterium in gastric mucosa but also induce innate and adaptive immune responses. Previous studies have revealed that individuals positive for H. pylori have a lower risk of some allergic diseases such as asthma [73]. H. pylori-induced immune responses mainly involve cell-mediated immunity, in which T helper cell 1 (Th1) was activated [74]. H. pylori-positive patients with tumorectomy had higher levels of Th1 cells, and a high level of Th1 is associated with a favourable outcome of GC [75].

Further evidence suggested that the H. pylori would be driven out of the gastric mucosa with the progression of H. pylori-related gastric atrophy and less production of acid by the atrophic mucosa [43, 76]. Some infected individuals are naturally eradicated, and the H. pylori rate decreases [42]. Moreover, H. pylori eradication has limited effects in reducing gastric cancer risk for those infected individuals who have already progressed to preneoplastic lesions as the biological change caused by H. pylori could not be reversible [31], which suggests that some H. pylori-related gastric cancer patients would be detected as H. pylori-negative when diagnosis. Our results suggest that patients positive for H. pylori are prone to have less tumor length (≤4.5 cm) and early-stage (T1) disease, although without statistical significance for the difference (Table 1). This trend has also been observed in other studies [77]. When conducting an analysis stratified by these factors, the association showed a tendency to be more prominent in less tumor length (≤4.5 cm) and less advanced T1-T2 stage (Fig. 2). Therefore, the magnitude of the effect of H. pylori on the prognosis of GC might be underestimated, as the “false negative” patients would confound the result to the direction of no association.

Three limitations should be addressed in our study. First, heterogeneity between included studies existed in the analysis. We used several strategies to account for the heterogeneity such as subgroup analysis and sensitivity analysis, however, we could not exclude the heterogeneity resulting from the differences in patient ethnicity, H. pylori detection method, and other possible non-identified factors. Second, we did not measure DFS, in our primary cohort study. We summarized the relevant studies in the meta-analysis and found that H. pylori exerted a protective role on DFS. Third, although a more exhaustive search strategy had been developed, we found out another 11 potential studies that only drew a conclusion or reported a P-value related to whether H. pylori infection could affect the prognosis of the GC population. These results were ineligible and could not be summarized in this meta-analysis.

Conclusions

In conclusion, the combined results from our primary cohort study and meta-analysis show that H. pylori-positive patients have a better prognosis compared with H. pylori-negative patients. More aggressive treatment and stringent surveillance strategies will be necessary for patients negative for H. pylori patients at cancer diagnosis.