Teeth Loss and Risk of Cancer at 14 Common Sites in Japanese

The oral cavity provides a gateway between the external environment and the gastrointestinal tract and acts in both food ingestion and digestion. Oral hygiene potentially affects gastrointestinal flora and nutritional status and may thus have implications for the development of chronic disease. Among the elderly, tooth loss, a persistent health problem that is caused mainly by periodontitis, develops over many years through the spread of bacterial infection and inflammation of the gum to the ligaments and bone that support the teeth (1).

The potential role of periodontal disease in chronic disease has been highlighted by recent epidemiologic findings of an association with the risk of ischemic heart disease (2, 3) and stroke (4). Periodontal disease has also been associated with an increased risk of cancer at several sites. Several studies have associated oral cancer with both tooth loss and poor oral hygiene independent of age, tobacco consumption, and alcohol consumption (5-7). Tooth loss has also been reported to be associated with a higher risk of esophageal (8, 9), gastric (9, 10), and pancreatic (11, 12) cancers. However, most of these studies have been conducted in western populations, and evidence from Asian populations is scarce.

Here, we comprehensively examined the relationship between tooth loss and the risk of 14 types of cancers in a large-scale, case-control study involving 5,240 Japanese cancer patients and 10,480 age- and sex-matched noncancer controls.

Cases and controls were selected from the database of the Hospital-based Epidemiologic Research Program at Aichi Cancer Center, conducted at Aichi Cancer Center Hospital. Details of the Hospital-based Epidemiologic Research Program at Aichi Cancer Center have been described elsewhere (13, 14). Briefly, patients were enrolled between January 2001 and November 2005. Information on lifestyle factors as well as other relevant factors, including the number of teeth lost, was collected from first-visit outpatients ages 18 to 79 years using a self-administered questionnaire. All exposures were measured with a self-administered questionnaire given at the time of first visit to our hospital and subsequently again before any diagnostic procedures were conducted. The number of teeth lost was determined by the question, “How many teeth do you have remaining?” and responses were categorized into four groups: group 1, number of remaining teeth, ≥ 21; group 2, 9 to 20; group 3, 1 to 8; and group 4, 0. Responses were checked by trained interviewers. Each patient was asked for information when healthy or before the current symptoms developed.

The rationale for the use of noncancer outpatients at our hospital as controls for epidemiologic studies to evaluate cancer risk factors is as follows. Like most general hospitals in Japan, Aichi Cancer Center accepts new outpatients who visit at their own discretion, with or without a doctor’s referral, and although called a cancer hospital, only 19% of all new outpatients have cancer (15). Among noncancer outpatients, 45% present with no abnormal findings on clinical examination and 35% with benign nonspecific diseases. Our previous study showed that the lifestyle patterns of first-visit outpatients “agree” with those in a randomly selected sample of the general population of Nagoya City (16). The questionnaire data were loaded into the Hospital-based Epidemiologic Research Program at Aichi Cancer Center database and periodically linked with the hospital-based cancer registry system to update the data on cancer incidence. The study was approved by the Institutional Ethics Committee at Aichi Cancer Center.

In the present study, patients ages 20 to 79 years with any of 14 types of incident cancer newly diagnosed from January 2001 to the end of 2005 were deemed eligible as case subjects. The 14 common cancers were defined according to the following codes of the International Classification of Diseases, Tenth Edition and Related Health Problems: head and neck (C00-06, 10-14, 30-32), esophagus (C15), stomach (C16), colorectum (C18-20), liver (C22), pancreas (C25), lung (C34), breast (C50), uterus (C53-55), ovary (C56), prostate (C61), bladder (C67), thyroid (C73), and lymphoma (C81-85). A total of 5,240 cancer patients with no prior history of cancer were enrolled. The controls were randomly selected from the 22,124 first-visit outpatients at Aichi Cancer Center Hospital enrolled during the same period in Hospital-based Epidemiologic Research Program at Aichi Cancer Center and confirmed on diagnostic investigation to have no cancer or history of neoplasia. Two age (±3 years)– and sex-matched controls were assigned to each case.

Smoking habit was entered under the three categories of never, former, and current. Cumulative smoking dose was evaluated as pack-years, the product of the number of packs consumed per day and years of smoking. Former smokers and drinkers were defined as those who quit smoking at least 1 year before the survey.

Consumption of food items, green-yellow vegetables, other vegetables, and fruits were based on the semiquantitative food frequency questionnaire, described elsewhere (17, 18), which consists of 47 single food items with frequencies in the eight categories of never or seldom, one to three times a month, one to two times a week, three to four times a week, five to six times a week, once a day, twice a day, and three or more times a day. Values for the consumption of these three food groups were calculated as the sums of consumption of contributing single food items as estimated from the food composition table, food frequency, and portion size. The food frequency questionnaire was validated by referring to a 3-day weighed dietary record as standard, which confirmed acceptable reproducibility and validity (19, 20).

We assessed the effect of tooth loss on the risk of cancer in terms of odds ratios (OR) and 95% confidence intervals (95% CI) as estimated by conditional logistic models adjusted for potential confounders. Potential confounders considered in the multivariate analyses were age, sex, smoking and drinking status (never, former, current), vegetable and fruit intake, body mass index (BMI; body weight/height²), and regular exercise (yes, no). Dietary intake was categorized into first, second, and third tertiles of intake among controls. Missing values for covariates were treated as a dummy variable and included in the models. Effect modifications were assessed by interaction terms between number of remaining teeth and sex, age, smoking status, or pack-years. Differences in categorized demographic variables between cases and controls were tested by the χ² test. Mean values for age and total dietary intake were compared between cases and controls by the Wilcoxon’s rank-sum test. P < 0.05 was used for statistical significance. All analyses were done using STATA version 8 (Stata).

Table 1 summarizes subject characteristics for the 14 common cancer sites and controls. A total of 5,240 cases and 10,480 age- and sex-matched controls were included. The proportion of current smokers and drinkers was higher in the subjects than in the controls. The proportion of current drinking tended to be higher in the subjects. The control group had a greater green-yellow vegetable intake than the case group, but no difference was seen in the amount of intake of other vegetables and fruits. The percentage of subjects undertaking regular exercise was “lower” in the control group.

Table 2 shows the association between tooth loss and the risk of cancer for 14 common site cancers. A decreased number of remaining teeth (increased number of teeth lost: number of remaining teeth, 0) was associated with increased OR of having head and neck (OR, 1.68; 95% CI, 0.88-1.93; P_trend = 0.055), esophagus (OR, 2.36; 95% CI, 1.17-4.75; P_trend = 0.002), and lung (OR, 1.54; 95% CI, 1.05-2.27; P_trend = 0.027) cancers compared with a preserved number of remaining teeth (number of remaining teeth, ≥21). In contrast, a decreased number of teeth remaining was also associated with lower OR for prostate cancer of 0.49 (95% CI, 0.19-1.26; P = 0.049). Drinking status had the biggest effect in pushing this association away from the null.

In view of the association of tooth loss with age, age-stratified analyses were conducted for head and neck, esophagus, lung, and prostate cancers (Table 3). For head and neck and esophageal cancers, clear associations between tooth loss and cancer risk were seen in the younger group (<70 years old; head and neck cancer OR, 2.98; 95% CI, 1.17-7.60; P_trend = 0.050; esophageal cancer OR, 4.43; 95% CI, 1.68-12.79; P_trend = 0.002; lung cancer OR, 1.94; 95% CI, 1.14-3.38; P_trend = 0.166). Associations were less clear in the older group (≥70 years old; head and neck cancer OR, 2.49; 95% CI, 0.95-6.53; P_trend = 0.079; esophageal cancer; OR, 1.68; 95% CI, 0.45-6.16; P_trend = 0.235).

Table 4 shows ORs for head and neck, esophageal, and lung cancers stratified by sex. In head and neck cancer, a positive association between tooth loss and OR was seen in women (OR, 3.44; 95% CI, 1.49-7.95; P_trend = 0.007) but not in men (OR, 1.14; 95% CI, 0.74-1.76; P_trend = 0.433). In esophageal cancer, a positive association was seen in both sexes but was stronger in women (OR, 4.16; 95% CI, 1.19-17.95; P_trend = 0.023) than men (OR, 1.65; 95% CI, 1.04-2.62; P_trend = 0.049). For lung cancer, in contrast, no association was evident in either sex. A statistically significant interaction between the number of remaining teeth and sex was evident in esophagus cancer (P = 0.023).

The effect of confounding by smoking habits was further evaluated in stratified analyses. The association of smoking with tooth loss is well known (21). Given that smoking status (never, former, current) and pack-years also affect the risk of various types of cancers, we stratified by smoking status and pack-years. Table 5 shows ORs for head and neck, esophageal, and lung cancers stratified by status and pack-years. Positive associations between tooth loss and ORs in never smokers (pack-years = 0) were seen for both head and neck (OR, 4.24; 95% CI, 1.56-11.49; P_trend = 0.008) and esophageal (OR, 9.50; 95% CI, 1.33-67.98; P_trend = 0.021) cancers but not in former or current smokers. In contrast, the association was obvious in current smokers for lung cancer. Statistically significant interactions between number of remaining teeth and pack-years were evident in head and neck cancer (P = 0.030).

Our case-control study shows that greater teeth loss was associated with an increased risk of head and neck, esophageal, and lung cancers and a decreased risk of prostate cancer. For head and neck and esophageal cancers, stratified analyses by sex or age showed clear associations between tooth loss and cancer risk in women and younger subjects but less clear associations in men and older subjects. In addition, this association was strongly observed in never smokers and never/former drinkers.

To date, associations with tooth loss have been reported for the risk of oral, esophageal, gastric, and pancreas cancers (5-12) but not for lung or prostate cancer. Our findings for head and neck and esophageal cancers are consistent with these previous epidemiologic studies. In contrast, we found no association with the risk of gastric or pancreas cancer as was also observed in two different cohorts study (10, 12). In addition, no previous study has reported a positive association with the risk of lung cancer or an inverse association with that of prostate cancer, although Hujoel et al. reported a positive association between periodontitis and lung cancer mortality (22). Our results therefore strengthen the evidence for an association with head and neck and esophageal cancers. The apparently discrepant results for other cancer sites require further clarification.

One of the central questions in previous publications examining tooth loss as a risk factor for cancer was whether residual confounding by smoking could explain significant associations. In the present study, analyses stratified by smoking status clearly showed a significant association between tooth loss and the risk of head and neck and esophageal cancers in never smokers. This finding strengthens the association between tooth loss and cancer risk observed in conventional analyses. Michaud et al. also reported a significant association between tooth loss and the risk of pancreas cancers in never smokers (12). However, other studies did not stratify by smoking status but did include smoking status in their models (e.g., pack-years or simply never, past, or current smoking).

Several mechanisms behind the association have been hypothesized. First, tooth loss is a common consequence of chronic bacterial infections, such as periodontitis (1). It may therefore serve as a surrogate for chronic infection and inflammation, which in turn may be important in the pathogenesis of cancer (23). Of interest, this mechanism may in particular support the associations with head and neck and esophageal cancers. In addition, the host response to periodontal disease may lead to systemic exposure to proinflammatory cytokines. In this way, periodontal disease may increase the risk of lung cancer through the chronic release of inflammatory mediators. Second, periodontal disease may influence carcinogenesis through the increased generation of carcinogens, specifically nitrosamines. Individuals with periodontal disease and poor oral hygiene have considerably elevated levels of oral bacteria and markedly higher nitrosamine levels in their oral cavity due to the presence of nitrate-reducing bacteria (24). Poor oral hygiene and the attendant increased loss of teeth might result in greater endogenous nitrosamine production and therefore a greater risk of these cancers. However, the possibility that tooth loss in our analyses merely reflects unhealthful behavior cannot be ruled out. Third, tooth loss reduces masticatory ability and may hence lead to the consumption of a less healthy diet, which could be associated with cancer (25-27). Reduction of masticatory ability also may result in the deglutition of larger pieces of food, leading to mechanical injuries of the mucous membrane in the oral cavity or esophagus.

A second interesting finding was the negative association between tooth loss and the risk of prostate cancer. Interpretation of this association is difficult. Several epidemiologic studies have shown a negative association between socioeconomic status and the risk of prostate cancer (28). Tooth loss is an indicator of poor oral hygiene, which may be associated with socioeconomic status. We speculate that this might be one reason for this inverse association. However, as we had no information on socioeconomic status in the current study, further study to clarify this point is warranted.

Several potential limitations of this study should be considered. First, the number of teeth remaining was examined after the occurrence of cancer. Although it is unlikely that the occurrence itself caused tooth loss, this limitation should nevertheless be considered when interpreting the results.

Second, recall bias is possible; because data regarding the number of remaining teeth were obtained from a self-reported questionnaire rather than oral examination, measurement error in reporting is possible. The difficulty in evaluating the accuracy of this self-reporting means that the possibility of misclassification cannot be excluded. In this regard, however, Douglass et al. reported a strong correlation between the self-reported number of teeth and the actual number determined on clinical examination (r = 0.97; ref. 29), and we therefore assumed that self-reporting had no adverse effect on our results. Nevertheless, this study was from a completely different population and might therefore not apply to our present Japanese population.

Third, residual confounding by smoking may be a major limitation. Although we adjusted for common confounders, including age, sex, smoking and drinking status (never, former, current), vegetable and fruit intake, BMI, and regular exercise, we cannot rule out residual confounding by smoking to explain the association we identified with lung cancer. Interestingly, the association between tooth loss and the risk of head and neck and esophageal cancers was obvious in never smokers. In contrast, this trend was not observed in lung cancer. The reason for this discrepancy is unclear. The strong association between smoking and both tooth loss and the risk of lung cancer is well known. Although smoking status was adjusted for as a potential confounder, the effect of smoking status might not have been excluded in the analyses, especially in lung cancer. The lack of information on the type of tobacco used (e.g., cigar, pipe, or chewing) may also be a problem. In addition, we could not adjust for other confounders, including infection with human papillomavirus, which increases the risk of head and neck cancers, or education. This is also a major limitation and a possible explanation for the positive findings. Further study to clarify this point is needed.

Fourth, a methodologic issue with the selection of the control base population remains. We used noncancer patients at our hospital as controls, given the likelihood that our cases arose within this population base. Moreover, earlier comparisons of lifestyle characteristics confirmed no substantial difference between these outpatients in our hospital and individuals randomly selected from the general population (16). The medical background of controls is another potential source of bias; however, our previous study, which focused on a female population, showed only a limited effect for this variable, and little difference would be expected for males.

In conclusion, our study has shown significant positive associations between tooth loss and the risk of head and neck, esophageal, and lung cancers, although the effect of smoking status might not have been sufficiently excluded, especially with regard to lung cancer. Our results suggest that the maintenance of tooth number, especially before older age, might prevent these cancers. These findings indicate that preventive efforts aimed at the preservation of teeth may decrease the risk of these cancers.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.