Periodontal Disease and Incidence of Cancers: A Systematic Review and Meta-analysis of Cohort Studies.

Introduction
According to the World Health Organization (WHO), cancer is the second leading cause of death worldwide.1 In 2022, there were nearly 20 million new cancer cases and 9.7 million cancer-related deaths globally. Lifetime risk estimates indicated one in five people will develop cancer during their lifetime, and one in nine men or one in twelve women will die from the disease.2 Additionally, a substantial proportion of patients face severe financial hardship due to cancer treatment costs, leading to concurrent economic stress and health crises.3,4
Periodontal disease (PD) is a common oral condition characterized by gingival inflammation and the destruction of periodontal tissues.5 It has emerged as a global public health concern,6,7 with a worldwide prevalence of up to 50%.8 Approximately 19% of the global adults suffered from severe PD, and the total number of cases worldwide exceeded one billion.9 PD has been reported to be linked to various systemic diseases,10, 11, 12 and its severe form can lead to tooth loss and edentulism.13,14
Emerging evidence suggested a potential association between PD and cancer risk. In 2017, a meta-analysis of 50 studies including both cohort and case-control designs identified a positive association between PD and oral, lung, and pancreatic cancers.15 Other meta-analyses focusing on specific cancer types have also demonstrated significant associations between periodontitis and head and neck16 or liver cancers.17 However, the association between PD and cancer risk remains controversial. A substantial body of studies18, 19, 20, 21, 22, 23, 24, 25, 26 reported positive associations, while a significant number of studies yielded null findings.27, 28, 29, 30, 31, 32 For instance, the Women’s Health Initiative Observational Study33 found an association between PD and the incidence of breast cancer, whereas no significant association was observed in the Sister Study.27 Additionally, a study published by the National Institutes of Health indicated that Fusobacterium nucleatum, a bacterium associated with PD, may promote the growth of colorectal cancer.34 Original articles published prior to 2017 were predominantly conducted in the United States and European countries, suggesting that previous meta-analyses on the PD-cancer association were primarily based on the Western populations, with limited representation from other regions.
Herein, we carried out an updated systematic review and meta-analysis of cohort studies to systematically evaluate the association between PD and multiple types of cancer.

Methods

Inclusion and exclusion criteria
Any cohort study that compared cancer incidence between participants with and without PD and met the inclusion criteria was eligible for this systematic review and meta-analysis. Inclusion criteria were as follows: 1) original research articles; 2) studies with a cohort or longitudinal design that report associations using hazard ratios (HR), relative risks (RR), or odds ratios (OR); 3) study subjects restricted to humans. The exclusion criteria were as follows: 1) studies deemed irrelevant by title or abstract (eg, case reports, reviews, case-control studies, cross-sectional studies, animal experiments); 2) fewer than two studies investigating the same type of cancer.

Data sources and searches
We systematically searched PubMed and Web of Science. The search-free text terms were (‘periodontal disease’ OR ‘periodontal disorders’) AND (‘cancer’ OR ‘tumor’ OR ‘neoplasm’ OR ‘sarcoma’ OR ‘carcinoma’ OR ‘malignant disease’ OR ‘malignant lesion’). To ensure the comprehensiveness of our search results, we supplemented this with targeted searches for each cancer type identified post hoc, using the following terms: ‘colorectal cancer’ OR ‘liver cancer’ OR ‘head and neck cancer’ OR ‘gastric cancer’ OR ‘lung cancer’ OR ‘hematologic cancer’ OR ‘kidney cancer’ OR ‘bladder cancer’ OR ‘pancreatic cancer’ OR ‘esophageal cancer’ OR ‘breast cancer’ OR ‘prostate cancer’. The search was restricted to English-language publications. Additionally, we systematically screened the reference lists of relevant primary studies and review articles to identify any additional eligible studies that might be missed in the initial database search.

Data extraction
Two investigators (Tureke M and DL) independently screened all titles and abstracts and retrieved full texts of potentially eligible studies. Using a duplicate and independent approach, the full texts were assessed against the inclusion criteria to determine eligibility. Any discrepancies were resolved through consensus, with consultation from the third investigator (CD). For all included studies, the following data were extracted: sample size, country or region, PD assessment methods, follow-up duration, and cancer types.
The Newcastle-Ottawa Scale (NOS)32 was used to assess the quality of the included studies. This scale comprises three primary domains: selection of study participants, comparability of groups, and ascertainment of outcomes. Two investigators (DL and Tureke M) independently evaluated the quality of the studies. Any disagreements in assessments were resolved through re-evaluation of the original article by the third investigator (CD). Studies were categorized based on their NOS scores: high-quality (scores of 8-9), medium-quality (scores of 6-7), and low-quality (scores ≤ 5).33

Data analysis
The primary objective of this study was to evaluate the association between PD and cancer risk in cohort studies. The extracted data were summarized. Incidence density was estimated by dividing the number of new cancer cases by the product of the sample size and geometric mean follow-up time, assuming uniform follow-up duration across participants. The effect size was assessed using RR and the corresponding 95% confidence interval (CI). Pooled estimates of OR, HR, or RR for each cancer type were calculated using either fixed-effects or random-effects models. Model selection was guided by the magnitude of heterogeneity, which was assessed using the I2 statistics: values of 30% or higher indicated moderate heterogeneity, while values exceeding 50% reflected substantial heterogeneity.35
Stratified analyses were conducted by PD assessment method (self-reported vs clinically examined), follow-up duration (short vs long, divided by the middle follow-up time of the included studies), and geographic region (restricted to America and Asian studies, European studies were excluded due to limited numbers). Sensitivity analyses were performed to verify the stability of results, using a leave-one-out approach to evaluate whether the exclusion of any single study significantly altered the overall findings. Publication bias were assessed using funnel plots and Egger’ tests.36,37 All statistical analyses were conducted using R software (version 4.4.1). All P-values were two-sided, with statistical significance defined as P < 0.05.
The study protocol was not registered with PROSPERO or another registry.

Results
A total of 8,816 studies were retrieved from PubMed and Web of Science. Following initial title and abstract screening, 8,761 irrelevant or ineligible studies were excluded, leaving 55 articles for full-text evaluation. After a detailed assessment of these articles, 31 studies were excluded due to the following reasons: three focused on mortality or other non-target outcomes; four had incomplete full texts; seven lacked effect size indicators; seven contained overlapping data across databases; nine examined tooth loss without addressing PD; one solely investigated the association between PD and ovarian cancer. Additionally, two relevant studies were identified through manual screening of the reference lists. Ultimately, 26 studies were included in the final analysis (Figure 1).
Figure 2 shows the distribution of studies by published year and geographical region. Studies of the PD and pancreatic cancer were pioneered in the United States in 2007 (51,529 participants). From 2022 to 2024, large cohort studies on the association between PD and bladder cancer were carried out in South Korea (713,201 participants in 2022; 200,170 in 2023). Substantial research on the association between PD and colorectal cancer was also conducted in South Korea. Long-term investigations into pancreatic cancer were conducted in Sweden.
Table 1 and Supplementary Table 1 summarize the characteristics of included studies. The analysis comprised a total of 16,257,141 participants and 53,884 cancer cases. The geometric mean follow-up duration per cancer type ranged from 10.03 years (for head and neck cancer) to 26.55 years (for prostate cancer). The majority of studies focused on the association between PD and colorectal cancer. Breast cancer exhibited the highest incidence density, at 3.31 × 10⁻3 person-years-1. Esophageal cancer studies had the largest sample size (6,177,678 participants), while melanoma studies had the smallest (49,712 participants). Studies investigating gastric cancer, liver cancer, and head and neck cancer predominantly enrolled populations from China and South Korea, whereas those focusing on breast cancer and prostate cancer predominantly included populations from the United States.
In terms of study quality, the NOS scores were generally high. Seventeen studies achieved NOS scores of eight or higher, while nine studies scored seven (Supplementary Table 2).
Positive associations were observed for multiple cancer types. The highest risk elevation (RR = 1.34, 95% CI: 1.04-1.71; I2
= 69.4%) were observed on esophageal cancer studies, followed by pancreatic cancer (RR = 1.26, 95% CI: 1.03-1.52; I² = 62.5%) and bladder cancer (RR = 1.21, 95% CI: 1.11-1.32; I² = 25.4%). Increased risk was also observed between PD and prostate cancer (RR = 1.17, 95% CI: 1.07-1.29; I² = 0.0%) and colorectal cancer showed (RR = 1.16, 95% CI: 1.01-1.34; I2 = 96.3%), with substantial heterogeneity. A consistent risk increase was found across gastric cancer studies (RR = 1.16, 95% CI: 1.08-1.25; I2 = 21.7%). A modest but statistically significant association was identified between PD and breast cancer (RR = 1.11, 95% CI: 1.03-1.17; I2 = 50.1%). For lung cancer, risk estimates were elevated and approached statistical significance (RR = 1.29, 95% CI: 0.97-1.72, I2 = 94.8%), constrained by marked heterogeneity. For hematologic cancer, the risk magnitude was comparable (RR = 1.32), but estimates were imprecise due to wide confidence intervals (95% CI: 0.84-2.09; I2 = 96.3%). Liver cancer showed a statistically significant association (RR = 1.12, 95% CI: 1.04-1.21; I2 = 0.0%), whereas kidney cancer exhibited a borderline non-significant trend (RR = 1.16, 95% CI: 0.93-1.44; I2 = 65.3%) (Figure 3; Supplementary Figure 1).
When stratified by PD assessment methods, self-reported PD was found to be positively associated with colorectal cancer (RR = 1.28, 95% CI: 1.01-1.62; I2 = 90.4%), gastric cancer (RR = 1.34, 95% CI: 1.11-1.63; I2 = 13.6%), lung cancer (RR = 1.31, 95% CI: 1.13-1.52; I2 = 0.0), bladder cancer (RR = 1.26, 95% CI: 1.12-1.43; I2 = 0.0%), and pancreatic cancer (RR = 1.42, 95% CI: 1.03-1.96; I2 = 68.6%). Exceptions included hematologic cancer (RR = 1.72, 95% CI: 0.98-3.02; I2 = 96.3%) and esophageal cancer (RR = 1.24, 95% CI: 0.98-1.57; I2 = 52.1%), for which no significant associations were observed. In contrast, clinically examined PD showed a positive association only with gastric cancer (RR = 1.11, 95% CI: 1.04-1.16; I2 = 0.0%). Contrasting but statistically non-significant effect sizes for hematologic cancer were found in self-reported (RR = 1.72, 95% CI: 0.98-3.02; I² = 0.0%) and clinically examined PD (RR = 0.97, 95% CI: 0.86-1.09; I² = 0.0%) (Figure 4; Supplementary Table 4). Stratified analyses by follow-up duration and geographic region revealed significant heterogeneities in the PD-cancer association (Supplementary Figures 4-5; Supplementary Tables 5-6). In stratification by follow-up duration, PD was associated with increased risk of breast cancer (RR = 1.16, 95% CI: 1.10-1.23; I2 = 0.0%) and head and neck cancer (RR = 1.20, 95% CI: 1.09-1.32; I2 = 0.0%) in the short follow-up group, whereas the association with esophageal cancer yielded a higher point estimate in the long follow-up group (RR = 1.35, 95% CI: 1.11-1.64; I2 = 0.0%) than in the short follow-up group (RR = 1.10, 95% CI: 1.05-1.17; I2 = 13.7%). Geographic stratification showed region-specific variations: the association between PD and gastric cancer was stronger in American group (RR = 1.42, 95% CI: 1.15-1.74; I2 = 0.0%) than in Asian group (RR = 1.10, 95% CI: 1.05-1.17; I2 = 0.0%). Similarly, the risk of hematologic cancer was significantly elevated in the American group (RR = 1.31, 95% CI: 1.13-1.53; I2 = 0.0%), but not in the Asian group (RR = 1.50, 95% CI: 0.65-3.46; I2 = 98.7%).
Results from Egger’ tests and funnel plots showed that no publication bias were identified in studies on lung, bladder, or prostate cancers. For colorectal, hematologic, and gastric cancers, Egger’ test showed no statistically significant bias, however, funnel plots suggested potential publication bias (Supplementary Figure 2). Leave-one-out sensitivity analysis revealed that excluding the study by Hu, J-M’ caused significant fluctuations in the pooled effect sizes for colorectal cancer. Similar variations were observed when removing the study by Mai Xiaodan (2016) for lung cancer and the study by Eun JK for hematologic cancer (Supplementary Figure 3).

Discussion
This meta-analysis of 26 cohort studies examined the association between PD and multiple cancer types. PD was found to have positive associations with colorectal, gastric, bladder, pancreatic, esophageal, breast, prostate, head and neck, and liver cancers. In studies with self-reported PD, increased risks of colorectal, pancreatic, gastric, bladder, and lung cancers were observed. In contrast, for studies using clinical examination to assess PD, only the risk of gastric cancer was significantly elevated.
Our meta-analysis found a positive association between PD and colorectal cancer, which was consistent with most prior research. In 2021, a meta-analysis of seven studies reported a 44% increased colorectal cancer risk in individuals with PD.38 Similarly, a meta-analysis of 11 studies reported a 21% increased risk in 2024.39 However, in 2017, a meta-analysis of five studies found no association between PD and colorectal cancer.15 This null result might be partially attributed to the small case numbers: two of the included studies had particularly low incident case counts (n = 6 and n = 17, respectively), which resulted in exceptionally wide confidence intervals (0.95-2.29) and limited statistical power. Furthermore, the association between PD and colorectal cancer exhibited high heterogeneity. Even after stratification by factors including PD assessment methods, follow-up duration, and geographic region, the pooled results continued to show substantial heterogeneity. This pattern may be attributable to residual confounding. Specifically, the number and scope of adjusted covariates appear to correlate with the magnitude of the effect estimates. For instance, studies that adjusted for only a limited set of basic demographic variables (eg, age, sex)-such as Xiaodan Mai (2016)-reported a higher effect size (HR = 2.09). In contrast, studies that incorporated a wide range of lifestyle, clinical comorbidity, and socioeconomic variables (eg, smoking, diet, diabetes, multiple comorbidity indices)-such as Eun Joo Kang (2023) -yielded effect estimates towards the null value (HR = 1.04). This suggests that inadequate adjustment for confounders may lead to upwardly biased effect estimates, thereby contributing to increased overall heterogeneity in the meta-analysis. Several biological mechanisms may contribute to the association. Chronic periodontal inflammation elevates systemic levels of pro-inflammatory cytokines, which may promote carcinogenesis and tumor progression.40,41 Additionally, oral pathogenic bacteria implicated in periodontitis (eg, Fusobacterium nucleatum and Porphyromonas gingivalis) are frequently detected in colorectal cancer and exhibit oncogenic properties.42,43
The positive association between lung cancer and PD in our study did not reach statistical significance, which differs from the findings of other meta-analyses.44, 45, 46 Our meta-analysis revealed substantial heterogeneity (I2 = 94.8%) in the association between PD and lung cancer. Stratification by diagnostic method showed that the association based on self-reported PD exhibited low heterogeneity (I2 = 29.7%), whereas the association of clinically examined PD demonstrated high heterogeneity (I2 = 98.2%). After stratified by region, subgroup heterogeneity in the American population was only 1.6%, suggesting that geographic region is related to overall heterogeneity. The low heterogeneity observed in the American subgroup may be attributed to greater consistency in population characteristics, exposure or outcome definitions, and adjustment for confounding factors. Further analysis indicated that the observed heterogeneity was primarily from methodological divergences in defining PD across studies: the 2024 South Korean study47 used administrative diagnostic criteria by combining ICD-10 classification (K05.2-K05.3) with periodontal procedure codes (eg, scaling or root planning) as proxies for clinical severity, while the 2022 South Korean study48 used radiographic quantification of alveolar bone loss via alveolar crest height (ACH) measurements, an objective but anatomically limited approach. Additionally, the 2017 United States study49 applied clinical assessment using the Community Periodontal Index (CPI≥3), which prioritizes probing depth over comprehensive staging. These distinct operationalizations of ‘PD’ all deviated from the gold-standard stage or grade system centered on clinical attachment loss, a measure constrained in practice by frequent challenges in accurately localizing the cementoenamel junction, leading to widespread reliance on PD despite known discrepancies with true attachment loss. While the radiographic (ACH) and clinical (CPI) methods were partially aligned with staging principles by quantifying tissue destruction, they capture non-identical pathological dimensions. ACH reflects historical bone loss, whereas CPI assesses current functional impairment. This difference accounts for their differential sensitivity to disease phenotypes relevant to systemic outcomes and contributes to the observed heterogeneity. Additionally, methodological heterogeneity in addressing confounders contributed to outcome variability. The 2017 United States study identified smoking as a significant effect modifier,49 revealing substantial variation in the periodontitis-lung cancer association across smoking strata. In contrast, the South Korean studies treated smoking solely as a covariate in overall models without assessing potential effect modification through stratified analyses.47,48 Consequently, the discrepancy between the South Korean and the United States estimates introduced methodological heterogeneity, as distinct smoking-subgroup associations were aggregated into single study-level effect measures.
Multiple meta-analyses support an association between PD and gastric cancer. For example, a meta-analysis of 19 studies in 202450 found that PD increased the risk of gastric cancer by 13%, an effect size smaller than in our study. Mechanistic evidence suggests that chronic inflammation and immune dysregulation triggered by PD may contribute to gastric cancer.51 However, despite consistent associations in epidemiological studies and meta-analyses, a bidirectional Mendelian randomization analysis failed to establish a causal relationship.52 Additionally, our findings demonstrate positive associations between PD and pancreatic cancer, head and neck cancer, and breast cancer risk, consistent with most of the previous studies.53, 54, 55, 56, 57, 58
Self-reported PD has been considered a feasible tool in epidemiological research. A Chinese study validated a high level of consistency between the self-report questionnaire and clinical assessment for the identification of periodontitis.59 Large epidemiological studies also demonstrated that self-report PD could be used for surveillance of periodontal disease.60 In our study, the association between PD and cancer showed slightly varying estimates by PD assessment. One potential explanation is that self-reported data reflect patients’ experiences of severe symptomatology, such as recurrent gingival bleeding, pain, tooth mobility, which aligns with long-term, clinically consequential periodontal destruction.61 However, self-reported PD assessment relies on participants’ subjective perception and recall ability, which may lead to misclassification bias – individuals with mild or asymptomatic PD may underreport their condition, while those with heightened health awareness may overreport. In contrast, clinical indices capture early or subclinical changes. Studies based on clinical examination metrics, which included transient or non-progressive cases, may thus have attenuated the associations compared to those using self-report data. Additionally, the stratum based on clinical diagnosis included fewer studies than the self-reported stratum, which could also contribute to the observed difference.
Our study included a large number of cohort studies from diverse regions, which strengthens the reliability of our findings. However, this study also had several limitations. First, significant heterogeneities were observed for some cancer types. The adjustment of covariates varied across studies. In the subgroup analysis, the reduction in heterogeneity may be attributable to the decreased number of studies available for analysis, rather than indicating true consistency. Secondly, despite the large sample size overall, the number of studies and participants for certain cancer types remained insufficient. Furthermore, this study did not explore the dose-response relationship between PD severity classification, and cancer risk. Finally, the search was restricted to two databases and English-language publications, which may introduce language bias.

Conclusions
In conclusion, this meta-analysis indicated that PD is associated with an increased risk of colorectal cancer, gastric cancer, bladder cancer, pancreatic cancer, esophageal cancer, breast cancer, and prostate cancer. Further studies exploring the associations between PD-related microbial alterations and cancer development, to provide novel insights into the mechanistic links between oral health and cancer, is warranted.

Author contribution
CD and Turek M conducted the statistical analysis. Turek M, DL, and CD performed literature search and data extraction. ZW contributed to study conception and design. CD and WW drafted the original manuscript. ZW, Turek M, LF, JZ, and CD revised the manuscript. All authors read and approved the final version of the manuscript.

Ethics statement
Not required because only (secondary) data from published studies were used.

Conflict of interest
The authors declare no conflict of interests