Study characteristics
The process used to select the studies included in this article is summarized in Figure 1. From an initial 1194 potentially relevant articles, the duplicate studies was removed and we screened titles and abstracts of articles. Finally, a total of 18 articles including 39 studies and 7264 participants were enrolled in the meta-analysis. The detailed characteristics of the selected studies are presented in Table 1. The selected articles were published from 2013 to 2018, and all articles were evaluated by the NOS (Supplementary Table 2). There were 9, 8, 4 and 3 articles related with incidence, OS, DFS, and CSS, respectively. 16 studies were conducted in Asia, 19 in North America and 3 in Europe. Among the 38 included studies, 10 studies involved patients with Pg infection, 15 with Fn infection, 5 with Tf infection, 3 with Aa infection, 3 with Td in infection and 2 with Pi infection. The sample sizes of the included studies ranged from 80 to 1069. According to the mean of all samples, 12 studies were considered to have a large sample size (n >467), while 11 had a small sample size (n ≤ 467).
Periodontal bacteria and incidence of cancer
23 studies with 10736 patients reported the relationship between periodontal bacteria and incidence of cancer (Figure 2). Periodontal bacteria infection increased the incidence of cancer as much as 1.25 times compared with those no infecting with periodontal bacteria (OR=1.25, 95%CI: 1.03–1.52, P =0.02) although with heterogeneity (I2=71%, Ph<0.00001). The subgroup studies consisted of different periodontal bacteria, ethnicity of participants , tumor location and sample size (Table 2). In our subgroup analysis, individuals whose infected with Pg were at 2.16 times greater risk of developing cancer than those no infecting with Pg (OR=2.16, 95%CI: 1.34–3.47; P =0.001; I2=79%, Ph <0.0001). Individuals with Pi infection exhibit increased incidence of cancer (OR=1.28; 95%CI: 1.01–1.63; P =0.04). However, there was no significant relation between the infection of Tf (OR=1.06, 95%CI: 0.8–1.41, P =0.67; I2=41%, Ph =0.15), Aa (OR=1.00, 95%CI: 0.48–2.08, P =1.00; I2=80%, Ph =0.006), Td (OR=1.30, 95%CI: 0.99–1.72, P =0.06; I2=6%, Ph =0.35 ), Fn (OR=0.61; 95%CI: 0.32–1.16; P =0.13; I2=0%, Ph =0.68) and incidence of cancer. There was association in Asia (OR=2.59, 95%CI:1.65–4.05, P <0.0001; I2=94%, Ph <0.00001) and Caucasian (OR=1.19, 95%CI:1.03–1.36, P=0.02; I2=44%, Ph=0.02) between periodontal bacteria infection and incidence of cancer. In the subgroup analysis of tumor location, incidence of cancer was associated with periodontal bacteria infection in OSCC (OR=10.02, 95%CI:3.11–32.33, P <0.0001) but not EC (OR=1.73, 95%CI:0.80–3.73, P=0.17; I2=82%, Ph=0.0007), PC (OR=1.21, 95%CI:0.96–1.53, P=0.1; I2=61%, Ph=0.02), PLGC (OR=0.69, 95%CI:0.37–1.29, P=0.24; I2=48%, Ph=0.12) and CRC (OR=1.26, 95%CI:1.00–1.57, P=0.05; I2=42%, Ph=0.12). According to the subgroup analysis of the sample size, periodontal bacteria infection were related to incidence of cancer in large sample size (OR=1.26, 95%CI:1.08–1.46, P=0.003; I2=46%, Ph=0.04), but not in small sample size (OR=1.24, 95%CI:0.73–2.12, P=0.42; I2=42%, Ph<0.00001).
Periodontal bacteria and OS in cancer
Figure 3a indicates that OS of cancer patients was evaluated in 8 studies with 3289 patients. The HR for OS in cancer patients infecting with periodontal bacteria compared with those no infecting with periodontal bacteria was 1.75 times (95% CI: 1.40–2.20, P <0.00001). The result revealed periodontal bacteria infection was related to poor OS in cancer. There was little heterogeneity between studies (I2=0%, Ph =0.79). The subgroup analysis involved in different periodontal bacteria (Mainly Pg and Fn), ethnicity of participants , tumor location and sample size (Table 3). In the subgroup, both of Pg and Fn infection was correlated with poor OS in cancer (Pg: HR=4.04, 95% CI: 1.54–10.63, P =0.05; Fn: HR=1.67, 95% CI: 1.32–2.11, P <0.0001). There was little heterogeneity between studies (Pg: I2=0%, Ph =0.64; Fn: I2=0%, Ph =0.99). Periodontal bacteria infection was correlated with poor OS of cancer patients in Asia (HR=1.90, 95%CI:1.43–2.53, P <0.0001; I2=0%, Ph=0.70) and Caucasian (HR=1.53, 95%CI:1.05–2.23, P=0.03; I2=0%, Ph=0.77). In the subgroup of tumor location, there were consistent findings in EC (HR=2.13, 95%CI:1.36–3.35, P=0.0010; I2=15%, Ph=0.31) and CRC (HR=1.64, 95%CI:1.26–2.13, P=0.0002; I2=0%, Ph=0.97). According to the subgroup analysis of sample size, periodontal bacteria infection exhibited a trend of correlation with poor OS in large (HR=1.53, 95%CI:1.05–2.23, P=0.03; I2=0%, Ph=0.77) and small sample size (HR=1.90, 95%CI:1.43–2.53, P<0.0001; I2=0%, Ph=0.70).
Periodontal bacteria and DFS in cancer
Figure 3b shows the results of DFS of cancer patients in 5 studies with 1384 patients. The HR for DFS in cancer patients infecting with periodontal bacteria compared with those not periodontal bacteria was 2.18 times (95% CI: 1.24–3.84, P=0.007). The result revealed there was significant association between periodontal bacteria infection and poor DFS in cancer. Interstudy heterogeneity was noted (I2 = 81%, Ph= 0.0003). The subgroup studies involved in types of periodontal bacteria, ethnicity of participants , tumor location and sample size (Table 3).
Periodontal bacteria and CSS in cancer
The association of periodontal bacteria infection and DFS in cancer was supplied by 3 studies with 1674 patients (Figure 3c). Data analysis showed that the periodontal bacteria infection was related to poor CSS (HR = 1.85, 95% CI: 1.44–2.39, P <0.00001) without obvious heterogeneity (I2 = 0%, Ph= 0.51).
Heterogeneity and sensitivity analysis
There was evidence of significant heterogeneity in incidence of cancer (I2=71%, Ph<0.00001) and DFS (I2 = 81%, Ph= 0.0003) but not OS (I2=0%, Ph =0.79) and CSS (I2 = 0%, Ph= 0.51). Subgroup analyses detecting potential sources of heterogeneity indicated that different periodontal bacteria, ethnicity of participants, tumor location and sample size were not significantly correlated with the heterogeneity in this meta-analysis. We found that Gao 2016 study was the source of heterogeneity in the meta-analysis for incidence of cancer and Oh 2018 for DFS. After removing Gao 2016 and Oh 2018, the heterogeneity among the studies decreased slightly for incidence of cancer (I2 = 64%, Ph<0.0001), but decreased significantly for DFS (I2 = 62%, Ph= 0.05), and the result for incidence of cancer (OR=1.21, 95%CI: 1.02–1.44, P=0.03) (Supplementary Figure 1) and DFS (HR=2.79, 95%CI: 1.72–4.54, P<0.0001) (Supplementary Figure 2) followed the same trends as those in the previous analysis. We also performed a sensitivity analysis through removing low-quality studies (NOS<7). The result for incidence of cancer (OR=1.29, 95%CI: 1.08–1.53, P =0.005) (Supplementary Figure 3), OS (OR=1.67, 95%CI: 1.28–2.18, P=0.0002) (Supplementary Figure 4), DFS (HR=2.34, 95%CI: 1.28–4.28, P=0.006) (Supplementary Figure 5) followed the same trends as those in the previous analysis.
Publication Bias
The results of the risk of bias assessment are shown in figure 4a, 4b. Egger’s and Begg’s tests indicated the potential publication bias for incidence (0.092) and prognosis (including OS, DFS and CSS: 0.624) of cancer. There was no significant publication bias in these studies.