In this prospective study of a cohort of 80-year-old Japanese subjects, we have shown that number of teeth lost might be an independent predictor of both all-cause and cancer mortality. Even after extensive adjustment for recognized confounders, tooth loss was significantly associated with an increased risk of all-cause and cancer mortality, but not of cardiovascular disease (CVD) and pneumonia mortality. Interestingly, when we further assessed the association between tooth loss and site-specific cancer mortality, there were significant associations between tooth loss and orodigestive cancers, but not between tooth loss and lung cancer.
To our knowledge, previous longitudinal studies investigating the association between tooth loss and orodigestive cancer are limited to reports by three groups: Stolzenberg-Solomon et al.
, Abnet et al.
[3, 4], and Michaud et al.
. The only investigation in an Asian population was by Abnet et al.
. Our study is the second Asian study and the first in a Japanese population, although a previous case–control study in 5,000 outpatients of a Japanese Cancer Center reported a positive association between tooth loss and the risk of head and neck, esophageal, and lung cancer
. In the study of Abnet et al.
[3, 4], there were two critical issues: 1) tooth loss was set as a dichotomous variable based on the median for the cohort, namely six teeth; and 2) only upper gastrointestinal (GI) cancers were investigated. In their study, upper GI cancer was defined as esophageal, gastric cardia, and non-cardia gastric cancers. In their report, tooth loss was significantly associated with increased risk of death from upper GI cancer (HR: 1.35, 95% CI: 1.14–1.59). Also, the risk of upper GI cancer associated with tooth loss was higher in male never-smokers than in male smokers (HR: 1.59, 95% CI: 1.03–2.45; HR: 1.39, 95% CI: 1.06–1.83, respectively). Results of multivariate analyses in a study in the United States
 found no significant associations between tooth loss and morbidity due to stomach cancer, pancreatic cancer, or colorectal cancer (HR: 1.10, 95% CI: 0.56–2.16; HR: 0.91, 95% CI: 0.56–1.47; HR: 1.10, 95% CI: 0.87–1.37, respectively). However, in another study by the same researchers, a significant association between tooth loss and pancreatic cancer morbidity was observed (HR: 1.61, 95% CI: 1.13–2.31)
. In another Western country, Finland, there was a significant association between tooth loss and pancreas cancer morbidity (HR: 1.63, 95% CI: 1.09–2.46)
. Thus, these studies have reported conflicting results for the association between tooth loss and cancer morbidity or mortality.
On the other hand, we found no significant association between tooth loss and CVD mortality, which is inconsistent with the recent other study, as reported by Watt et al.
. This difference may be due to the difference in subject age (the mean age of their study was approximately 50 years old in that study).
One of the possible reasons underlying the different outcomes could be because of differences in the criteria used for measuring tooth loss. In most cases, a dichotomous variable was used. For example, Stolzenberg-Solomon et al.
 used two categories, edentulism vs. 0–10 missing teeth, while Michaud et al.
 used three categories of 0–16, 17–24, and 25–32 teeth. Those results indicate that cut-off values for number of teeth utilized there have not been standardized. Because of this disparity, we employed the number of missing teeth as a continuous variable in the present study, as it basically represents an accumulated burden of severe periodontal disease as the number increases. However, the association might not always be linear, as the association between CVD mortality and the number of missing teeth was shown to be non-linear in the study of Tu et al.
. A similar situation has been also been found in periodontal disease assessments. Periodontal disease is generally diagnosed by probing and its diagnosis of periodontal disease is not straightforward, thus it would be inappropriate to use for determination of the underlying disease status. As pointed out by Tu and Gilthorpe
, an alternative method is to use the number of lost teeth, as tooth loss appears to be a better indicator than probing as a marker of lifetime oral health, and is less prone to measurement error. International standardization regarding evaluation by tooth loss or periodontal disease is required.
There are several limitations in the present study. First, the sample consisted largely of generally healthy elderly subjects, who might have been more eager and/or able to participate. Thus, our findings may indicate an association only in generally healthy elderly subjects. Second, our subjects in this study were all 80 years old at baseline, which is a very elderly population. It is possible that the mortality rates for both total and specific cancers could be underestimated because of a survivor effect. Thus, future investigations in a younger population will be necessary to confirm the validity of our results. Third, evaluation of the mortality risk for specific cancers was limited because of the small number of site-specific orodigestive cancer deaths.