Association of frailty index with incident complete tooth loss among middle-aged and older Chinese adults: a longitudinal cohort study based on CHARLS

Abstract

Background

Tooth loss is a severe oral health condition among middle-aged and older adults and is associated with impaired mastication, nutritional problems, and reduced quality of life. Frailty—a state of increased vulnerability to adverse health outcomes—may represent an underexplored factor associated with oral health outcomes, but population-based evidence from China remains limited. This study aimed to evaluate the association between the 35-item Frailty Index (FI35) and incident complete tooth loss using data from the China Health and Retirement Longitudinal Study (CHARLS).

Methods

A longitudinal cohort analysis using data from the 2015 and 2018 waves of the China Health and Retirement Longitudinal Study was conducted. Frailty and covariates were assessed at baseline in 2015, and incident complete tooth loss was assessed at follow-up in 2018. Participants who met the definition of complete tooth loss at baseline were excluded. Multivariable logistic regression was used to assess the association between FI35 (as a continuous variable and in quartiles) and incident complete tooth loss, with sequential adjustment for demographic characteristics, lifestyle factors, and comorbidities. Restricted cubic splines were employed to explore nonlinear dose–response patterns. Subgroup analyses and interaction tests were applied to examine effect modification.

Results

During the three-year follow-up, 603 (6.03%) participants developed incident complete tooth loss. In the fully adjusted model, Each 0.1-unit increase in baseline FI35 was associated with 16% higher odds of incident complete tooth loss (OR = 1.16, 95% CI: 1.07–1.25, P < 0.001). Compared with participants in the lowest quartile, those in the highest quartile had 78% higher odds of incident complete tooth loss (OR = 1.78, 95% CI: 1.35–2.33; P for trend < 0.001). Restricted cubic spline analysis suggested a nonlinear association (P for overall = 0.002, P for nonlinear = 0.034). Exploratory subgroup analyses suggested possible heterogeneity by age (P for interaction = 0.032), although this finding should be interpreted cautiously.

Conclusion

Higher baseline FI35 was associated with greater odds of incident complete tooth loss during the three-year follow-up among middle-aged and older Chinese adults. Further studies incorporating objective oral-health assessments are needed to confirm these findings and clarify the underlying pathways.

Introduction

Tooth loss is one of the most prevalent oral health conditions among middle-aged and older adults worldwide. Recent data from the Global Burden of Disease Study 2021 indicate that complete tooth loss affects over 350 million people globally, with prevalence increasing sharply after age 60 [1]. In China, the burden of oral diseases, including tooth loss, has been rising substantially over the past three decades [2]. Beyond its immediate impact on masticatory function and nutrition, complete tooth loss has been consistently associated with a higher risk of cardiovascular disease, diabetes, cognitive impairment, and all-cause mortality [3]. Identifying factors associated with complete tooth loss remains an important public health priority.

Frailty is a multidimensional clinical state characterized by increased vulnerability to adverse health outcomes, resulting from age-related declines in physiological reserve across multiple organ systems [4]. The frailty index (FI), originally developed by Rockwood and colleagues, quantifies the accumulation of health deficits and has been widely validated as a predictor of mortality, hospitalization, and disability in older populations [5, 6]. In China, frailty has emerged as a notable public health concern, with prevalence estimates ranging from 10% to 30% among community-dwelling older adults [7].

Emerging evidence suggests that frailty may be associated with poor oral health outcomes [8]. Frail individuals often exhibit impaired physical function, reduced manual dexterity, and cognitive decline, all of which may compromise daily oral hygiene practices [9]. Additionally, frailty is characterized by chronic low-grade inflammation (sometimes termed “inflammaging”) and dysregulated immune responses, which could accelerate periodontal tissue destruction and alveolar bone resorption [10]. Such mechanisms provide a potential biological basis for a link between frailty and complete tooth loss.

Despite the plausibility of this association, epidemiological evidence remains limited. Most studies have focused on the relationship between frailty and oral health in Western populations [11, 12], whereas the association between frailty and complete tooth loss has not been systematically examined in large, nationally representative Chinese cohorts. The China Health and Retirement Longitudinal Study (CHARLS) offers a unique opportunity to address this gap, as it contains comprehensive health deficit data enabling the construction of a validated frailty index [13].

The present longitudinal cohort study used data from the 2015 and 2018 waves of CHARLS to examine whether baseline FI35 was associated with incident complete tooth loss during follow-up among Chinese adults aged 45 years and older. We further explored whether this association follows a linear or nonlinear dose–response pattern and whether it varies across population subgroups defined by demographic and behavioral characteristics.

Methods

Study population and data source

Data were drawn from CHARLS, a nationally representative survey of adults aged 45 years and older in China. The study uses a multistage stratified probability sampling design covering 28 provinces, municipalities, and autonomous regions. Detailed descriptions of the sampling strategy and data collection have been published previously [13]. We conducted a longitudinal cohort analysis using data from the 2015 and 2018 waves of CHARLS. The 2015 wave was considered the baseline assessment, at which FI35 and all covariates were measured. Tooth-loss status was reassessed during the 2018 follow-up.

Among 19,712 participants aged 45 years or older in the 2015 wave, 5,021 participants who reported complete tooth loss at baseline were excluded. Of the remaining 14,691 participants, six had missing information on complete tooth loss at the 2018 follow-up. Among the 14,685 participants with complete exposure and outcome data, 4,685 had missing covariate data and were excluded from the primary complete-case analysis, leaving 10,000 participants in the final analytical sample. The participant selection process is shown in Fig. 1.

Fig. 1.

Flowchart of participant selection

Frailty index (FI35) assessment

Frailty was quantified using a 35-item deficit accumulation frailty index (FI35), constructed following the standard methodology proposed by Rockwood and colleagues [5, 14]. The FI35 included deficits across several domains, including activities of daily living, instrumental activities of daily living, physical functional limitations, selected chronic diseases, mental health, and self-rated health. Deficits were coded on a scale ranging from 0 to 1, with higher values indicating greater severity. Binary disease variables were coded as 0 or 1, whereas ordinal functional and psychological variables were assigned intermediate values according to the severity or frequency of the deficit. The FI35 score was calculated by summing the individual deficit scores and dividing the total by the number of available deficits, resulting in a score ranging from 0 to 1. Higher scores indicate greater frailty. This approach has been previously applied and validated in Chinese populations [15].

Outcome definition: complete tooth loss

The primary outcome was incident complete tooth loss between the 2015 baseline and the 2018 follow-up. At both survey waves, participants were asked the following question from the general health status and disease history section: “Have you lost all of your teeth?” Responses were recorded as “yes” or “no.” Complete tooth loss was defined as a response of “yes” to this question. Participants who reported complete tooth loss at baseline in 2015 were excluded from the longitudinal analytical cohort. Incident complete tooth loss was defined as reporting no complete tooth loss at baseline in 2015 and reporting complete loss of all natural teeth at the 2018 follow-up [16].

Covariates

Covariates were selected a priori based on their established or plausible associations with both frailty and complete tooth loss [17]. All covariates were assessed at baseline in 2015. Sociodemographic covariates included age (continuous), sex (male or female), residence (rural or urban), marital status (married or non-married), and education (illiterate, junior high school, or high school or above). Anthropometric and lifestyle covariates included body mass index (kg/m²), smoking status (never or ever), and drinking frequency (never, less than once per month, or more than once per month). Clinical covariates included hypertension, diabetes mellitus, dyslipidemia, and cardiovascular disease, each categorized as yes or no.

Statistical analysis

All analyses were performed with R version 4.2.0. Baseline characteristics were compared between participants with and without complete tooth loss using Student’s t-test (for continuous variables) or the chi-square test (for categorical variables).

We used multivariable logistic regression to estimate the association between FI35 and complete tooth loss. Three logistic regression models were fitted. The crude model was unadjusted. Model 1 was adjusted for age, sex, residence, marital status, and education. Model 2 was adjusted for age, sex, residence, marital status, education, body mass index, smoking status, drinking frequency, hypertension, diabetes mellitus, dyslipidemia, and cardiovascular disease. Because cardiovascular disease was also included as a component of FI35, a sensitivity analysis excluding cardiovascular disease from the Model 2 covariate set was performed to assess the potential influence of overlapping adjustment.

FI35 was examined both as a continuous variable and in quartiles. For the continuous analyses, FI35 was multiplied by 10; therefore, the reported odds ratios represent the change in the odds of incident complete tooth loss associated with each 0.1-unit increase in the original FI35 score. For the categorical analysis, participants were classified into quartiles according to the baseline FI35 distribution, with the lowest quartile serving as the reference group [18]. To explore potential nonlinearity, we fitted a fully adjusted logistic regression model using restricted cubic splines for FI35 on its original 0–1 scale. Four knots were placed at the 5th, 35th, 65th, and 95th percentiles of the FI35 distribution (0.0094, 0.0620, 0.1280, and 0.3189, respectively). The median FI35 value of 0.0906 was used as the reference, and odds ratios with 95% confidence intervals were estimated relative to this value. The overall association and nonlinear component were assessed separately. Exploratory subgroup analyses were conducted according to age (45–60 vs. ≥60 years), sex, marital status, education, residence, smoking status, drinking frequency, hypertension, diabetes mellitus, dyslipidemia, and cardiovascular disease. Subgroup analyses were based on Model 2, with the corresponding stratification variable omitted from the adjustment set. Interaction between FI35 and each subgroup variable was assessed using likelihood ratio tests. Because multiple subgroup and interaction tests were performed, these analyses were considered exploratory, and nominal P values for interaction were interpreted cautiously. All tests were two-sided, and a P-value < 0.05 was considered statistically significant.

Results

Participant characteristics

After excluding participants with complete tooth loss at baseline and those with missing exposure, outcome, or covariate data, 10,000 participants were included in the final complete-case analysis. The participant flow diagram is presented in Fig. 1.

Baseline characteristics stratified by complete tooth loss status are shown in Table 1. The mean age of the total sample was 60.24 ± 8.69 years, and 603 participants (6.03%) had incident complete tooth loss. Compared with those without complete tooth loss, participants with complete tooth loss were older (65.6 vs. 59.9 years, P < 0.001), had lower BMI (23.22 vs. 24.14 kg/m², P < 0.001), and had higher FI35 (0.15 vs. 0.12, P < 0.001). They were also more likely to be female, non-married, have lower education, and live in rural areas (all P < 0.01). No significant differences were observed for smoking, drinking, hypertension, diabetes, dyslipidemia, or cardiovascular disease, although the complete tooth loss group had slightly higher proportions of hypertension and diabetes.

Table 1.

Baseline characteristics of study participants according to incident complete tooth-loss status during follow-up

Characteristic	Total(n = 10000)	No(n = 9397)	Yes(n = 603)	P value
Age	60.24 ± 8.69	59.89 ± 8.57	65.61 ± 8.77	< 0.001
BMI	24.08 ± 3.80	24.14 ± 3.79	23.22 ± 3.78	< 0.001
FI35	0.12 ± 0.10	0.12 ± 0.10	0.15 ± 0.11	< 0.001
Sex				0.16
female	5372(53.72)	5031(53.54)	341(56.55)
male	4628(46.28)	4366(46.46)	262(43.45)
Marital status				< 0.001
Married	8848(88.48)	8371(89.08)	477(79.10)
Non-Married	1152(11.52)	1026(10.92)	126(20.90)
Education				< 0.001
High school or above	1171(11.71)	1137(12.10)	34(5.64)
Illiterate	2400(24.00)	2174(23.14)	226(37.48)
Junior high school	6429(64.29)	6086(64.77)	343(56.88)
Residence place				< 0.001
rural	6372(63.72)	5947(63.29)	425(70.48)
urban	3628(36.28)	3450(36.71)	178(29.52)
Smoke				0.45
no	7273(72.73)	6826(72.64)	447(74.13)
yes	2727(27.27)	2571(27.36)	156(25.87)
Drink				0.09
<1 m	862(8.62)	808(8.60)	54(8.96)
>1 m	2657(26.57)	2520(26.82)	137(22.72)
no	6481(64.81)	6069(64.58)	412(68.33)
Hypertension				0.06
no	5669(56.69)	5350(56.93)	319(52.90)
yes	4331(43.31)	4047(43.07)	284(47.10)
Diabetes mellitus				0.33
no	8280(82.80)	7790(82.90)	490(81.26)
yes	1720(17.20)	1607(17.10)	113(18.74)
Dyslipidemia				0.10
no	6529(65.29)	6116(65.08)	413(68.49)
yes	3471(34.71)	3281(34.92)	190(31.51)
Cardiovascular disease				0.26
no	8095(80.95)	7618(81.07)	477(79.10)
yes	1905(19.05)	1779(18.93)	126(20.90)

Table 1 Baseline characteristics of participants according to incident complete tooth loss status during follow-up.

Association between FI35 and complete tooth loss

The multivariable logistic regression results are presented in Table 2. In the crude model, each 0.1-unit increase in baseline FI35 was associated with 30% higher odds of incident complete tooth loss (OR = 1.30, 95% CI: 1.21–1.39, P < 0.001). After adjustment for age, sex, residence, marital status, and education in Model 1, the association remained statistically significant (OR = 1.12, 95% CI: 1.04–1.21, P = 0.004). In the fully adjusted Model 2, each 0.1-unit increase in baseline FI35 was associated with 16% higher odds of incident complete tooth loss (OR = 1.16, 95% CI: 1.07–1.25, P < 0.001).

Table 2.

Association between baseline FI35 and incident complete tooth loss

Characteristic	crude model			Model 1		Model 2
	OR (95% CI)		P	OR (95% CI)	P		OR (95% CI)	P
FI35	1.3(1.211,1.394)		< 0.001	1.12(1.037,1.21)	0.004		1.16(1.065,1.25)	< 0.001
FI35 quartiles
Q1	Reference			Reference			Reference
Q2	1.425(1.088,1.867)		0.001	1.24(0.946,1.637)	0.124		1.27(0.962,1.66)	0.096
Q3	1.825(1.405,2.372)		< 0.001	1.42(1.083,1.853)	0.011		1.49(1.132,1.95)	0.004
Q4	2.47(1.930,3.162)		< 0.001	1.62(1.248,2.099)	< 0.001		1.78(1.354,2.33)	< 0.001
P for trend			< 0.001		< 0.001			< 0.001

Crude model: unadjusted

Model 1: adjusted for age, sex, residence, marital status, and education

Model 2: adjusted for variables in Model 1 plus body mass index, smoking status, drinking frequency, hypertension, diabetes mellitus, dyslipidemia, and cardiovascular disease.Model 2: adjusted for variables in Model 1 plus body mass index, smoking status, drinking frequency, hypertension, diabetes mellitus, dyslipidemia, and cardiovascular disease

FI35 quartiles were defined as Q1, FI35 ≤ 0.043; Q2, 0.043 < FI35 ≤ 0.091; Q3, 0.091 < FI35 ≤ 0.157; and Q4, FI35 > 0.157

When FI35 was analyzed in quartiles, a significant positive trend was observed across increasing FI35 categories. In the fully adjusted model, compared with participants in the lowest quartile (Q1), the odds ratios for incident complete tooth loss were 1.27 (95% CI: 0.96–1.66, P = 0.096) for Q2, 1.49 (95% CI: 1.13–1.95, P = 0.004) for Q3, and 1.78 (95% CI: 1.35–2.33, P < 0.001) for Q4. The estimates for Q3 and Q4 were statistically significant, whereas the estimate for Q2 did not reach statistical significance. The trend across FI35 quartiles was statistically significant (P for trend < 0.001). In the sensitivity analysis excluding cardiovascular disease from the Model 2 adjustment set, the direction and magnitude of the associations remained materially consistent with those of the primary analysis.

Restricted cubic spline analysis

In the fully adjusted restricted cubic spline model, baseline FI35 was associated with incident complete tooth loss (P for overall association = 0.002; P for nonlinearity = 0.034). FI35 was modeled on its original scale, with the median value (0.0906) used as the reference. Knots were placed at the 5th, 35th, 65th, and 95th percentiles of FI35 (0.0094, 0.0620, 0.1280, and 0.3189, respectively). The spline curve suggested a possible nonlinear pattern, with higher FI35 generally associated with greater odds of incident complete tooth loss. However, the distribution of FI35 showed that relatively few participants had high FI35 values, and the confidence intervals widened in the upper range.

Subgroup analyses

Exploratory subgroup analyses are presented in Fig. 2, together with subgroup-specific sample sizes and numbers of incident complete tooth-loss events. All odds ratios represent each 0.1-unit increase in the original FI35 score.

Fig. 2.

Subgroup analyses of the association between baseline FI35 and incident complete tooth loss

Nominal evidence of interaction was observed for age (P for interaction = 0.032). The association was stronger among participants aged 45–60 years (OR = 1.33, 95% CI: 1.10–1.61) than among those aged 60 years or older (OR = 1.10, 95% CI: 1.00–1.20). No statistically significant interactions were observed for sex, marital status, education, residence, smoking status, drinking frequency, hypertension, diabetes mellitus, dyslipidemia, or cardiovascular disease (all P for interaction > 0.05). Given the multiple interaction tests, these subgroup findings should be interpreted as exploratory.

Discussion

In this longitudinal cohort of Chinese adults aged 45 years and older, higher baseline FI35 was associated with greater odds of incident complete tooth loss during the three-year follow-up. The association persisted after adjustment for demographic, lifestyle, and clinical covariates and showed a nonlinear pattern. Exploratory subgroup analyses suggested possible heterogeneity across age strata; however, this finding requires cautious interpretation.

Complete tooth loss is the end result of a complex interplay between dental diseases, including caries and periodontitis, systemic conditions, and behavioral factors [19, 20]. Over the past decade, frailty has become a key concept in geriatric medicine, capturing cumulative decline across multiple physiological systems [21]. Frail individuals are at increased risk of a range of adverse health outcomes, including falls, hospitalization, and mortality [22]. Our findings extend this literature by demonstrating a longitudinal association between frailty and incident complete tooth loss. However, the potential clinical relevance of this association requires further evaluation in studies incorporating objective oral-health assessments.

Several hypothetical pathways may underlie the observed association. Previous studies have shown that frailty is often accompanied by chronic low-grade inflammation, sometimes referred to as “inflammaging” [23]. Elevated levels of pro-inflammatory cytokines, including tumor necrosis factor alpha (TNF-α), interleukin-6 (IL-6), and interleukin-1 beta (IL-1β), have been reported in frail individuals [24], and these cytokines are involved in periodontal inflammation, osteoclast differentiation, and alveolar bone resorption [25]. Frailty has also been associated with impaired immune function, reduced physical capacity and manual dexterity, nutritional deficiencies, and sarcopenia [26–29]. These factors could plausibly contribute to poorer oral health through increased susceptibility to oral disease, reduced ability to maintain oral hygiene, or impaired tissue maintenance. However, the present study did not assess periodontal status, oral-hygiene practices, inflammatory biomarkers, alveolar bone changes, or oral microbiome characteristics. These pathways should therefore be regarded as hypotheses based on previous literature rather than mechanisms demonstrated by the present findings.

The restricted cubic spline analysis suggested a nonlinear association between FI35 and incident complete tooth loss, as illustrated in Fig. 3. The odds of incident complete tooth loss increased more steeply at lower FI35 levels and subsequently tended to plateau or decline slightly at higher levels. This pattern is broadly consistent with observations from studies of other frailty-related outcomes, in which the relationship between deficit accumulation and adverse events may be nonlinear [30]. However, the confidence intervals widened considerably at higher FI35 levels because relatively few participants were represented in this range. Therefore, the apparent plateau or decline should be interpreted cautiously and should not be regarded as evidence of a definitive threshold or a protective effect at higher FI35 levels.

Fig. 3.

Restricted cubic spline analysis of the association between baseline FI35 and incident complete tooth loss. The solid line represents the odds ratio and the shaded area represents the 95% confidence interval. FI35 was modeled on its original scale. The reference value was the median FI35 (0.0906). Knots were placed at the 5th, 35th, 65th, and 95th percentiles of FI35 (0.0094, 0.0620, 0.1280, and 0.3189). The lower panel shows the distribution of FI35 in the analytic sample

Exploratory subgroup analyses suggested that the association may differ across age strata. However, the subgroup-specific estimates were less precise than the overall estimate, and multiple interaction tests were conducted. Differences in event numbers across strata, sampling variability, or other age-related factors may have contributed to this finding [31, 32]. Therefore, the apparent age-related heterogeneity should be considered hypothesis-generating and requires confirmation in independent cohorts.

Several limitations should be considered when interpreting these findings. First, although the longitudinal design established the temporal ordering of baseline FI35 and subsequent complete tooth loss, the observational nature of the study precludes causal inference. Residual confounding and bidirectional relationships between frailty and underlying oral-health deterioration cannot be completely excluded [33]. Second, FI35 was constructed partly from self-reported health deficits and may therefore be affected by reporting bias, although similar frailty-index approaches have been validated in Chinese populations [34]. Third, complete tooth loss was self-reported and was assessed using a binary questionnaire item; therefore, information on the number of remaining teeth, timing and causes of tooth loss, dental caries, periodontal status, and tooth extraction was unavailable [35]. Fourth, CHARLS did not provide detailed information on oral-hygiene behavior, dental-care utilization, dietary intake, inflammatory biomarkers, alveolar bone status, or oral microbiome characteristics. Consequently, the proposed mechanisms could not be directly examined, and residual confounding by these factors remains possible [36]. However, the multiple-imputation analysis produced results consistent with the complete-case analysis. Finally, the study was limited to Chinese adults aged 45 years or older, which may restrict the generalizability of the findings to other populations [37].

Future studies should incorporate repeated assessments of frailty together with objective dental examinations, including the number of remaining teeth, periodontal status, dental caries, reasons for tooth extraction, and oral-health behaviors. Studies incorporating dietary factors, healthcare utilization, inflammatory biomarkers, and oral microbiome characteristics may help clarify the biological and behavioral pathways underlying the observed association. Whether interventions addressing frailty can reduce the occurrence of complete tooth loss remains uncertain and requires evaluation in appropriately designed prospective and interventional studies.

Conclusion

In this longitudinal cohort of middle-aged and older Chinese adults, higher baseline FI35 was associated with greater odds of incident complete tooth loss during the three-year follow-up. The association showed evidence of nonlinearity, although estimates at higher FI35 levels were less precise. Further studies incorporating objective dental examinations and detailed oral-health-related factors are needed to confirm these findings and clarify the underlying pathways.

Authors' contributions

CQ and WZ designed the research. XL collected, analyzed the data. LG drafted the manuscript. revised the manuscript. All authors contributed to the article and approved the submitted version.

Funding

This work did not receive any specific grant.

Data availability

The data that support the findings of this study are available from the China Health and Retirement Longitudinal Study (CHARLS) repository. Access to these data can be obtained by registering and submitting a request through the official CHARLS website at http://charls.pku.edu.cn.

Declarations

Ethics approval and consent to participate

The present study was a secondary analysis of publicly available, de-identified data from the China Health and Retirement Longitudinal Study (CHARLS). The original CHARLS study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Peking University (IRB00001052–11015). Written informed consent was obtained from all participants by the CHARLS research team. As the present study used only publicly available, de-identified secondary data and did not involve new participant recruitment, direct contact with participants, or collection of biological samples, additional ethical approval from the authors’ institution was not required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.