Tooth loss and cancer risk: a dose–response meta analysis of prospective cohort studies

Conflicting results to identify the relationship between tooth loss and cancer risk. Therefore, a dose-response meta-analysis was performed to clarify and quantitative assessed the correlation between tooth loss and cancer risk. Up to March 2017, 25 observational epidemiological studies were included in current meta-analysis. Tooth loss was significantly associated with a higher risk of cancer. Additionally, tooth loss was associated with significantly a higher risk of esophageal cancer, gastric cancer, head and neck cancer, colorectal cancer, pancreas cancer, lung cancer, prostate cancer, bladder cancer and hematopoietic cancer. Subgroup analysis showed consistent findings. Furthermore, a significant dose-response relationship was observed between tooth loss and cancer risk. Increasing per 10 of tooth loss was associated with a 9% increment of cancer risk, 14% increment of esophageal cancer risk, 9% increment of gastric cancer risk, 31% increment of head and neck cancer risk, 4% increment of colorectal cancer risk, 7% increment of pancreas cancer risk, 19% increment of lung cancer risk, 2% increment of bladder cancer risk and 3% increment of hematopoietic cancer risk. Considering these promising results, tooth loss might be harmful for health. Large sample size, different ethnic population and different cancer type are warranted to validate this association.


INTRODUCTION
Cancer has to be the second-leading cause in 2015 that caused over 8.8 million deaths worldwide in 2015 [1]. Due to countries geographical environment, living habits, cancer incidence is very different. The incidence of cancer in developing countries such as Africa, Asia and South America is the most severe. A total of 14 million cancer cases were added worldwide and 8.2 million people died in 2012. Among them, China added 3.07 million cancer patients and caused about 2.2 million deaths, accounting for 21.9% of the total global deaths. The incidence of cancer in developed countries is still higher than in developing countries. Residents of North America and Europe are the most vulnerable to cancer, but China has a large population base, making it the country with the highest number of cancer deaths throughout the world. As more and more people in developing countries improve their living standards and dietary patterns change, the chances of cancer have increased significantly in developing countries [2]. The etiology of cancer involves both genetic and environmental factors. Therefore, understanding the impact of environmental factors on cancer will help to prevent cancer.
Oral cavity is an important part of the body, and is starts in the digestive system, mainly by the lip and cheek, tongue and palate, salivary glands, teeth and jaw, with mastication, swallowing, speech and feeling, and

Meta-Analysis
Oncotarget 15091 www.oncotarget.com other functions, which maintain the normal shape of maxillofacial. Oral health is an important part of human health. The World Health Organization (WHO) identifies dental health as one of the top ten criteria for human health. Poor oral health may increase systemic inflammation, resulting in a local overly aggressive immune response, and thus could have important implications for cancer development. Periodontal disease and tooth loss are two common oral health measures [3]. Tooth loss has been considered to impact quality of life [4], and been known to considerably influence food choice, diet, nutrition intake, and esthetics [5].
Previous studies have examined the correlation between tooth loss and cancer risk . However, the result remains controversial. Additionally, no study to quantitative assessed tooth loss in relation to cancer risk. Thus, we performed this dose-response meta-analysis to clarify and quantitative assessed the correlation between tooth loss and cancer risk.

MATERIALS AND METHODS
This meta-analysis was conducted according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) checklist [31].

Search strategy
We included eligible studies to investigate the relationship between tooth loss and cancer risk in general adult populations. To develop a flexible, non-linear, r meta-regression model, we required that an eligible study should have categorized into 3 or more levels.
PubMed and EMBASE were searched for studies that contained risk estimates for the outcomes of cancer and were published update to March 2017, with keywords including "dentition" [

Study selection
Two independent researchers investigate information the correlation between tooth loss and cancer risk: outcome was cancer. Moreover, we precluded nonhuman studies, reviews, meta-analyses, editorials and published letters.

Data extraction
Use standardized data collection tables to extract data. Each eligible article information was extracted by two independent researchers. We extracted the following information: first author; publication year; age; country; sex; cases and participants; the categories of tooth loss; relative risk or odds ratio (OR). We collect the risk estimates with multivariable-adjusted [32]. According to the Newcastle-Ottawa scale, quality assessment was performed for non-randomized studies [33]. The disagreements were resolved through consensus by all the authors.

Statistical analysis
We pooled relative risk estimates to measure the association between tooth loss and cancer; the hazard ratio were considered equivalent to the relative risk [34]. Results in different subgroups of tooth loss and cancer risk were treated as two separate reports.
Due to different definitions cut-off points in the included studies for categories, we performed a relative risk estimates by the method recommended by Greenland, Longnecker and Orsini and colleagues [35]. Dose of tooth loss used the median tooth loss. If the median tooth loss category was not available, the midpoint of the upper and lower boundaries was considered the dose of each category. In addition, using restricted cubic splines to evaluate the non-linear association between tooth loss and cancer risk, with three knots at the 10th, 50th, and 90th percentiles of the distribution. A flexible meta-regression based on restricted cubic spline (RCS) function was used to fit the potential non-linear trend, and generalized least-square method was used to estimate the parameters. This procedure treats tooth loss (continuous data) as an independent variable and logRR of diseases as a dependent variable, with both tails of the curve restricted to linear. A P value is calculated for linear or non-linear by testing the null hypothesis that the coefficient of the second spline is equal to zero [32].
The between-study heterogeneity was assessed by Q-statistic (signifcance level at P ≤ 0.10) and the I 2statistic. STATA software 14.0 (STATA Corp, College Station, TX, USA) was using in all analyses. P < 0.05 was considered signifcant for all tests.

Literature search results
We identifed 3088 relevant citations after removing duplicates. Reviewing their titles and abstracts, 3021 citations were excluded. The remaining 67 citations were assessed in more detail for eligibility by reading the full text. Among them, 43 were excluded, after review reference, one articles was included. Finally, 25 studies were used for the final data synthesis . The flow chart of literature searching was presented in Figure 1, and the data were extracted. These studies were published update to March 2017. www.oncotarget.com

Tooth loss and overall cancer risk
Eighty-one independent reports from twenty five studies investigated the association between tooth loss and cancer . Compared with the lowest tooth loss, tooth loss is significantly associated with a higher risk of cancer risk (RR:1.06; 95% CI, 1.02-1.09; P < 0.001) ( Table 2). Additionally, a dose-response analysis revealed that each 10 tooth loss was associated with a 9% incremental in cancer risk (RR: 1.09; 95% CI, 1.05-1.13) ( Figure 2). A cubic spline model revealed an positive non-linear correlation between tooth loss and cancer (P < 0.001 for non-linearity; Figure 2).

Tooth loss and gastric cancer
Nine independent reports from five studies investigated the association between tooth loss and gastric   (Figure 3). A cubic spline model revealed an positive non-linear correlation between tooth loss and gastric cancer (P < 0.001 for non-linearity; Figure 3).

Tooth loss and pancreas cancer
Five independent reports from five studies investigated the association between tooth loss and pancreas cancer [12-14, 21, 30]. Compared with the lowest tooth loss, tooth loss is significantly associated with a higher risk of pancreas cancer (RR:1.15; 95% CI, 1.05-1.19; P < 0.001) ( Table 2). Additionally, a doseresponse analysis revealed that each 10 tooth loss was associated with a 7% incremental in pancreas cancer risk (RR:1.07; 95% CI, 1.01-1.15; P < 0.001) ( Figure  4). A cubic spline model revealed an positive non-linear correlation between tooth loss and pancreas cancer (P < 0.001 for non-linearity; Figure 4).

Tooth loss and lung cancer
Five independent reports from two studies investigated the association between tooth loss and lung cancer [12,13]. Compared with lowest tooth loss, tooth loss is significantly associated with a higher risk of lung cancer (RR: 1.66; 95% CI, 1.34-1.97; P < 0.001) ( Table 2). Additionally, a dose-response analysis revealed that each 10 tooth loss was associated with a 19% incremental in Oncotarget 15098 www.oncotarget.com lung cancer risk (RR:1.19; 95% CI, 1.04-1.35; P < 0.001) (Figure 4). A cubic spline model revealed an positive nonlinear correlation between tooth loss and lung cancer (P < 0.001 for non-linearity; Figure 4).

Tooth loss and bladder cancer
Two independent reports from two studies investigated the association between tooth loss and bladder cancer [12,13]. Compared with lowest tooth loss, tooth loss is significantly associated with a higher risk of bladder cancer (RR: 1.23; 95% CI, 1.12-1.35; P < 0.001) ( Table 2). Additionally, a dose-response analysis revealed that each 10 tooth loss was associated with a 2% incremental in bladder cancer risk (RR: 1.02; 95% CI, 1.01-1.03; P < 0.001) (Figure 4). A cubic spline model revealed an positive non-linear correlation between tooth loss and bladder cancer (P < 0.001 for non-linearity; Figure 4).

Tooth loss and hematopoietic cancer
Two independent reports from two studies investigated the association between tooth loss and hematopoietic cancer [12,13]. Compared with the lowest tooth loss, tooth loss is significantly associated with a higher risk of hematopoietic cancer (RR: 1.07; 95% CI, 1.02-1.13; P < 0.001) ( Table 2). Additionally, a doseresponse analysis revealed that each 10 tooth loss was associated with a 3% incremental in hematopoietic cancer risk (RR: 1.03; 95% CI, 1.01-1.07; P < 0.001) ( Figure  4). A cubic spline model revealed an positive non-linear correlation between tooth loss and hematopoietic cancer (P < 0.001 for non-linearity; Figure 4).

Subgroup analyses
Subgroup analysis was performed to check the stability of the primary outcome. Subgroup meta-analyses in study design and study quality showed consistent findings ( Table 2).

Sensitivity analysis
Sensitivity analysis was conducted to assess the stability of the results. The results show the results were stable in Supplementary Figure 1.

Publication bias
Each studies in this meta-analysis were performed to evaluate the publication bias by both Begg's funnel plot and Egger's test. P > 0.05 was considered no publication bias. The results show no obvious evidence of publication bias was found in the associations between tooth loss and cancer risk (Supplementary Table 2).

DISCUSSION
Cancer affects millions of people in developed and developing countries that is now a public health crisis. Despite the decline in the mortality rate of developed countries, cancer is still the main cause of death and has caused serious social and economic distress on a global scale over the past few decades. In low and middle-income countries, the incidence of cancer has risen sharply [2]. To date, there are a few identified risk factors for acute pancreatitis, including smoking, infection, occupational exposure, environmental pollution, unreasonable diet and genetic factors. Previous studies indicated that tooth loss may be a risk factors in cancer, but presented controversial results.
In the current meta-analysis was based on 25 case-control or cohort study, with more than 1.3 million participants and 32925 cases from eleven countries. Thus, this meta analysis provides the most up-to-date epidemiological evidence supporting tooth loss is harmful for cancer. A dose-response analysis revealed that a per 10 of tooth loss increase was associated with a 9% increment of cancer risk, 14% increment of esophageal cancer risk, 9% increment of gastric cancer risk, 31% increment of head and neck cancer risk, 4% increment of colorectal cancer risk, 7% increment of pancreas cancer risk, 19% increment of lung cancer risk, 3% increment of prostate cancer risk, 2% increment of bladder cancer risk and 3% increment of hematopoietic cancer risk. Subgroup meta-analyses by various factors also showed consistent findings.
Several plausible pathways may reasonable for the relationship between tooth loss and cancer. The influence of chronic inflammation on cancer development is one possible pathway. Chronic systemic inflammation linked to periodontal disease [36,37], which is a major cause of tooth loss in adults that can increase the risk of cancer by inhibiting apoptosis and stimulating tumor cell proliferation [38]. Secondly, the main cause of teeth loss is dental caries [39,40], and carbohydrate intake is the dental caries cause. Carbohydrate intake was associated with increased risk cancer [41,42]. Third, the progress of tooth damage destroys normal periodontal tissue, allowing oral microbial accumulation deep into oral tissue, thereby promoting its growth [43]. Thus, tooth loss and cancer seems to be closely related.
To our knowledge, this is the first study to identify and quantify the potential dose-response association between tooth loss and cancer risk in a large cohort of both men and women. Although, we performed this meta-analysis very carefully, however, some limitations must be considered in the current meta-analysis. First, different sex of population should be included in this meta-analysis to explore the impact of different sex of population on tooth loss and cancer risk. Second, we only select literature that written by English, which may have resulted in a language or cultural www.oncotarget.com bias, other language should be chosen in the further. Third, in the subgroup analysis in cancer type, there might be insufficient statistical power to check an association. Fourth, Though some data were obtained from retrospective studies, most of retrospective studies only to identify the relationship between tooth loss and cancer risk, and we also add more study to identify and quantify the potential dose-response association between tooth loss and cancer risk.
In conclusion, our findings underscore the notion that tooth loss was associated with cancer risk increment. In the future, large-scale and population based association studies must be performed in the future to validate the risk identified in the current meta-analysis.

CONFLICTS OF INTEREST
Authors have disclosed no conflicts of interest.

FINANCIAL SUPPORT
This work was partially supported by the Initial Project for Post-Graduates of HuBei University of Medicine (K1228901).