**3. Results**

#### *3.1. Characteristics of the Subjects Who Participated in the Study*

The study population consisted of 138 men and 205 women, who were all aged 85 in 2002. After 15 years, 12 subjects had survived. Their health status was evaluated by blood tests. The results are shown in Supplementary Table S1.

#### *3.2. Structure of QOL, ADL, and Self-Assessed Chewing Ability of the Older Adults*

#### 3.2.1. Structure of QOL

The SF 36 consists of eight subscales. Descriptive statistics of the eight subscales are presented in Supplementary Table S2. For these subscales, factor analysis was carried out through the major factor method with varimax rotation. Factor scores were used as summary scores of the factors for the following analysis. The results are shown in Supplementary Table S3. The subscales consisted of two factors. These factors were named the function and the role. Based on this result, structural equation modeling (SEM) was carried out. The results are shown in Figure 1. Body pain (BP) and physical functioning (PF) correlated with both latent variables.

**Figure 1.** Structure of quality of life of the older subjects at the age of 85. The subscales of the SF 36 consisted of two latent variables, named Function and Role. All paths were statistically significant. BP and PF correlated with both latent variables. Subscales: Physical functioning (PF), Role physical (RP), Body pain (BP), General health (GH), Vitality (VT), Social functioning (SF), Role emotional (RE), Mental health (MH), e: Error variable. SF 36: 36-Item Short-Form Health Survey. REMSEA: root-mean-square error of approximation.

#### 3.2.2. Structure of ADLs

The TIMG index consists of three subscales and has a total of 13 items. The scores of these items and the descriptive statistics of the three subscales are shown in Supplementary Tables S4 and S5. For the structure of the IADL, factor analysis and SEM were carried out in the same way as for the QOL. The results of the factor analysis are presented in Supplementary Table S6. Factor scores were used as summary scores of the factors for the following analysis. The model with SEM is shown in Figure 2. Items of the TIMG Index involved three factors. Visiting sick friends and filling out the pension form were correlated with two latent variables. Correlations between latent variables were statistically significant. However, the correlations were very weak.

**Figure 2.** Structure of the ADL. Items of the TIMG Index involved 3 factors. Visiting sick friends and filling out the pension. Pension were correlated with two latent variables. Correlations between latent variables were statistically significant. However, the correlations were very weak. ADL: activity of daily living, TMIG index: The Tokyo Metropolitan Institute of Gerontology index of competence. e: Error variable. REMSEA: root-mean-square error of approximation

Serum Albumin (g/dL)

Self-assessed

 Chewing ability

#### 3.2.3. Structure of Self-Assessed Chewing Ability and Correlation with Number of Remaining Teeth

Self-assessed chewing ability was evaluated by whether participants were able to chew 15 foods. The variables were dichotomous. To calculate the summary score, the item response theory analysis (IRT) was carried out. The item response curve and item information curves are shown in Figure S1, and the model is shown in Supplementary Table S7. Similar to the QOL and IADL, a factor analysis was carried out for these 15 foods. The results are shown in Supplementary Table S8. The fifteen foods had three factors, and the factors were named easy to chew food, slightly hard to chew food, and moderate and hard to chew foods. The correlations among chewing ability, number of remaining teeth, and serum level of albumin as indicators of nutritional status were analyzed by SEM (Figure 3). The link between chewing ability and serum albumin was not statistically significant (*p* = 0.692). Other than that, all associations were statistically significant.

**Figure 3.** Correlations among number of remaining teeth, chewing ability, and serum level of albumin. REMSEA: root-mean-square error of approximation

#### *3.3. Interaction of Nutritional Status, and Self-Assessed Chewing Ability with IADL and QOL*

#### 3.3.1. Correlation between Self-Assessed Chewing Ability and QOL

 0.106

> 0.336

A generalized liner model was applied to the dimensions of QOL calculated by the factor scores presented in Section 3.2.1. The results were shown in Table 1. For both function and role, the summary score of self-assessed chewing ability was significantly correlated. However, the number of remaining teeth and serum level of albumin were not statistically significant.

**QOL Function Role Coe**ffi**cient (95% CI)** *p***-Value Coe**ffi**cient (95% CI)** *p***-Value** Intercept −0.411 (−1.839–1.018) 0.573 −0.204 (−1.671–1.262) 0.785 Number of remaining teeth −0.002 (−0.017–0.014) 0.842 −0.006 (−0.022–0.009) 0.437

(−0.243–0.454)

(0.204–0.469)

**Table 1.** Correlations of the number of remaining teeth, serum albumin level, and self-assessed chewing ability with quality of life (QOL).

The generalized liner model was applied for the factor scores of QOL. Distribution: Normal, Link: Normal. The SF-36 consisted of eight subscales. For these subscales, the generalized linear model was applied. The results are presented in Table S9. Self-assessed chewing ability had a statistically significant correlation with PF, RP, GH, BT, and M, but not with BP, SF, or RE. The number of remaining teeth and serum level of Albumin had no correlations with the eight subscales. CI: confidence interval.

 0.553

<0.001  0.060

> 0.135 (0–0.271)

(−0.298–0.418)

 0.742

 0.050

#### 3.3.2. Correlation between Self-Assessed Chewing Ability and IADL

The TIMIG Index consists of three subscales: Self-management, Intercultural activity, and Social role. For these subscales, the generalized liner model was applied. The results were shown in Table 2. The summary score of self-assessed chewing ability calculated by IRT was significantly correlated with the three subscales. However, the number of remaining teeth and the serum level of albumin were not statistically significantly correlated with this factor.

**Table 2.** Correlations of the number of remaining teeth, serum albumin, and self-assessed chewing ability with the IADL.


The generalized liner model was applied to the factor scores of the TMIG Index. Distribution: Normal, Link: Normal. TMIG index: Tokyo Metropolitan Institute of Gerontology index of competence. IADL: instrumental activity of daily living. CI: confidence interval.

#### *3.4. E*ff*ects of Nutritional Status, Self-Asssessed Chewing Ability, and IADL on Mortality*

To assess the mortality rate at the 15-year follow up, nutritional status evaluated by the serum level of albumin, subscales of self-assessed chewing ability, and the IADL were analyzed using Cox's proportional hazard model. The results are presented in Table 3.


**Table 3.** Hazard ratios of nutritional status, self-assessed chewing ability, and IADL.

TMIG index: The Tokyo Metropolitan Institute of Gerontology index of competence. CI: confidence interval.

For women, only serum albumin level was shown to have a statistically significant effect on mortality, and its hazard ratio was the highest. In contrast, for men, the self-assessed chewing ability of moderate hard food, and intercellular activity had statistically significant effects on mortality. The number of remaining teeth did not have a statistically significant effect. However, when classified as edentulous or dentate, the hazard ratio of edentulous was statistically significant in men (hazard ratio: 1.766, 95% CI; 1.119–2.788, *p* = 0.015). Additionally, hazard ratios of chewing ability were adjusted by health status evaluated by blood tests. Results were shown in Supplementary Table S10. For men, adjusted hazard ratios of self-assessed chewing ability were statistically significant except for Creatinine.

The Kaplan–Meier analysis was used to calculate the survival rate. As self-assessed chewing ability is a contentious variable, the ability to chew three foods (konnyaku jelly, tubular roll of boiled fish paste, and steamed rice) was used as a dichotomous variable. The means and medians of the survival rate are shown in Supplementary Table S11. The survival curves of the statistically significant factors are shown in Figure 4.

**Figure 4.** Survival curves of the significant factors for mortality. (**A**) Serum levels of albumin for women. (**B**) Ability to chew Konnyaku-jelly of men. (**C**) Ability to chew Tubular roll of boiled fish paste of men. (**D**) Ability to chew Steamed rice of men. (**E**) Edentulous. (**F**)Intellectual activity of men.

#### *3.5. Overview of the Interactions Among Health-Related Factors*

Finally, by using all health-related factors investigated in this study, multiple group structural equation modeling was conducted for men and women. The results are presented in Figure 5. Black lines indicated statistical significance for both men and women, blue lines indicate significance only in men, red lines indicate significance only in women, and orange lines indicated no significance for either men or women. Self-assessed chewing ability was not associated with serum albumin. ADLs were not associated with QOL.

**Figure 5.** Overview of the interactions among health-related factors. Black lines indicated statistical significance for both men and women, blue lines indicate significance only in men, red lines indicate significance only in women, and orange lines indicated no significance for either men or women. e: Error variable. QOL: quality of live. ADL: activity of daily living. REMSEA: root-mean-square error of approximation.

## **4. Discussion**

In this study, nutritional status, evaluated by the serum level of albumin, was associated with mortality in women. Self-assessed chewing ability was significantly associated with quality of life (QOL) and the instrumental activity of daily living (IADL) evaluated by the TIMG Index.

The subjects who participated in this study were functionally independent and could attend mass check-ups held at the local city hall or gymnasium. No subjects were hospitalized or living in a nursing home. According to the Kaplan–Meier analysis, their mean life expectancy was 91.28 years for men and 94.38 years for women (Supplementary Table S6). The subjects who participated in this study represented a healthy and long-living population. A previous report showed a large difference in mortality between participants and non-participants in health check-ups [51]. The results of this study may not applicable for hospitalized older adults or older adults residing in nursing homes.

Several studies have suggested that regular diet [52] and nutritional status [53–57] affect the QOL of community-dwelling older adults. Another study showed that chewing ability is significantly greater in subjects with high QOL scores. Dietary intake was not associated with QOL [58]. In this study, chewing ability was significantly associated with two dimensions of QOL. However, nutritional status, as evaluated by the serum level of albumin, and number of remaining teeth were not directly associated with QOL. As shown in Figure 3, the number of remaining teeth is a morphological background factor in oral function [59,60]. Therefore, the number of remaining teeth is not directly associated with QOL. Nutritional status was evaluated by the serum albumin level, which is one of the limitations of this study. A more precise evaluation of nutritional status or regular diet by a validated questionnaire may lead to more precise results. However, these tools were not available when the survey was conducted in 2002.

In this study, the self-assessed chewing ability was associated with three subscales of the TMIG index. The number of remaining teeth and the serum level of albumin were not associated with the IADL. A previous report showed that tooth loss is associated with future decline in higher-level functional capacity [61]. Tooth loss can be compensated for by prosthodontic treatment. In addition, the Japanese national insurance system covers most conventional prosthodontic treatments. Recently, the concept of functional teeth was introduced, and it could be used as a predictor of mortality instead of the number of remaining teeth [62]. One of the limitations of this study is that we did not have data on functional teeth. However, there were only three out of 196 (1.5%) edentulous subjects who did not use dentures.

Recently, the concept of frailty, including oral frailty, has been widely accepted [63–65]. Frailty has been evaluated by physical conditions that can be improved by nutritional interventions. For nutritional intervention studies, frail is a more optimal outcome variable than ADLs [66–70]. ADLs do not only describe limited physical conditions. They include other dimensions such as social function and intellectual activity [71]. A previous study showed that physical activity, social role, and mental health are associated with ADLs [41,43,72]. This may be one of the reasons why the serum level of albumin was not significantly associated with the ADL subscales.

Nutritional factors a ffect mortality in older adults [73]. In this study, malnutrition was evaluated by the serum level of albumin [74]. A low level of serum albumin is a well-known predictor of mortality in older persons in both the short and long term [74–80]. The results of this study are consistent with another report conducted in women; however, it was not applicable in men. Except for one subject, all women with less than the cut-o ff point of albumin died within the observational period. They died within 2000 days. In contrast, one man was alive after the observational period and he became a centenarian. When men and women were combined, the hazard ratio for the serum level of albumin was 1.979 (95% CI: 1.172–3.341, *p* = 0.011).

Self-assessed chewing ability was significantly correlated with QOL and mortality in men. The number of remaining teeth was not statistically significantly correlated with mortality. The subjective method for the evaluation of chewing ability requires a specific device, labor, and costs. Due to its ease of use and cost e ffectiveness, masticatory dysfunction has generally been assessed by self-assessment-specific questionnaires in epidemiological studies. Studies have shown that the mortality of older adults is influenced by the number of remaining teeth. However, they failed to show a dose–response relationship [81,82]. As mentioned for the QOL, this may be because the e ffects of tooth loss can be compensated for by the use of proper dentures. In this study, the number of remaining teeth did not directly influence mortality. However, for edentulous subjects, not using dentures was significantly high risk in men (Hazard ratio: 15.160 (*p* = 0.019)). Therefore, tooth loss should be used in combination with the use of dentures [83,84]. Therefore, the concept of functional teeth is reasonable [62]. However, complete denture wearers and subjects with all-natural teeth were treated as equivalent. Further study is necessary to apply the concept of functional teeth in epidemiological studies. The number of remaining teeth should be considered as one of the indicators of oral function. Mortality is a multifactorial issue, and some related factors cause either tooth loss or mortality. In particular, socioeconomic status and health literacy may be important factors in mortality. In this study, we could not obtain these data. It is one of the limitations of this study. However, the association of self-assessed chewing ability with mortality rather than the number of teeth was a reasonable result. Hazard ratio of self-assessed chewing ability for slight hard food was statistically significant for men. It was also significant adjusted by blood tests except for creatinine. Creatinine reflects the muscle activity. Self-assessed chewing ability may reflect the exercise in daily life. However, as shown in Figure 4, clear survival curves were obtained. Self-assessed chewing ability for slight hard food can be the indicator for the prediction of mortality.

There is a sex difference in mortality related to the number of remaining teeth [46,85–88]. Most studies have shown that tooth loss is a risk factor for mortality in men and not in women [85,87,88]. Other studies have shown contradictory results [46,86]. One report had statistically not significant between men and women [89]. Follow-up periods, the baseline number of remaining teeth, and statistical methods were different between studies. In addition, mortality is a multifactorial issue, and some related factors cause either tooth loss or mortality. Prevalence of noncommunicable diseases may be different between studies. Health care supply system varies from country to country. In addition, the prevalence of noncommunicable diseases may be different between men and women. It is impossible to reach a clear conclusion for the sex difference of mortality.

Figure 5 shows one of the models. The interactions between health-related factors are different between men and women. When planning a health promotion plan for older adults, different strategies may be necessary for men and women.
