Type 2 diabetes mellitus and 20 year incidence of periodontitis and tooth loss

Abstract

Aims

The objective of this study was to evaluate the prospective associations between type 2 diabetes mellitus (T2DM) and the risk of periodontitis and tooth loss.

Methods

35,247 male participants of the Health Professionals Follow-Up Study who were dentate, free of periodontitis and cancer at baseline, were followed from 1986-2006. Data on self-reported diabetes, periodontitis, tooth loss and potential confounders were collected at baseline and biennially through mailed questionnaires. The multivariable adjusted relationships between diabetes and first report of periodontitis and tooth loss were estimated using time-varying Cox models.

Results

There were 3,009 incident self-reported periodontitis and 10,017 tooth loss events over 591,941 person-years. Men with T2DM showed a 29% (HR=1.29; 95% CI:1.13-1.47) increased risk of periodontitis compared to those without, when adjusted for age, race, smoking, BMI, fruit and vegetable intake, physical activity, alcohol consumption and dental profession. Men with T2DM with total fruit and vegetable intake <median were 49% as likely to report incident periodontitis compared to those without T2DM (HR=1.49; 95% CI: 1.23-1.80; p-value for interaction=0.03). The multivariable adjusted risk of tooth loss was 1.10 (95% CI: 1.02-1.18).

Conclusions

Type 2 diabetes mellitus was associated with a significantly greater risk of self-reported periodontitis.

Introduction

Diabetes affects an estimated 25.6 million people, nearly 11.3% of the U.S. population, including individuals with diagnosed and undiagnosed diabetes [2]. While micro- and macrovascular complications of diabetes such as retinopathy, neuropathy, nephropathy and cardiovascular disease are well recognized, the Centers for Disease Control and Prevention and National Institute of Diabetes and Digestive and Kidney Diseases also identified periodontitis as another important complication [2-4]. It has been widely accepted in the dental literature that type 2 diabetes mellitus (T2DM) is a risk factor for periodontitis [5] with biologic plausibility to support the underlying mechanisms, yet only 4 prospective studies to date have been published evaluating the association. Three of these have been conducted in the same unique population of Pima Indians [6-8] and another [9] compared the progression of periodontitis among already periodontally compromised subjects. Moreover, all studies were limited by a short follow-up time and adjustment of confounders. The available evidence has shown an increased risk of incidence and severity of disease for both type 1 and 2 diabetes [10]. Given the high prevalence of periodontitis, the public health impact of an association between diabetes and periodontitis would be substantial among those affected by diabetes. Periodontitis has been associated with cardiovascular outcomes and the subsequent tooth loss may result in deleterious changes in diet placing those with diabetes at even greater risk of co-morbid conditions. Therefore, we evaluated the association between T2DM and risk of self-reported periodontitis and tooth loss independent of important confounders in the Health Professionals Follow-up Study (HPFS), an ongoing closed cohort.

Subjects, Materials and Methods

Study Population

The HPFS consists of 51,529 US male health professionals aged 40-75 years at baseline who completed a mailed questionnaire in 1986 to evaluate associations between diet, heart disease and cancer; details have been previously described [11]. The initial questionnaire collected data regarding diet, lifestyle behaviors, anthropometric measures, medication use, medical and dental history. Follow-up questionnaires have been mailed biennially since 1988 and a detailed semi-quantitative food frequency (FFQ) questionnaire has been mailed every 4 years. Over 90% of the baseline population has responded to follow-up questionnaires [11].

Participants were excluded at baseline if they reported a history of periodontitis, answered don’t know or had missing information, were edentulous, had missing data on age, reported a history of cancer (excluding non-melanoma skin cancer), or provided poor dietary data resulting in a final sample of 35,247 men.

We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during this research. Approval by Harvard School of Public Health Board for Protection of Human Subjects Institutional Review Board was obtained.

Exposure Assessment

Self-reported diabetes status was collected from each biennial questionnaire with the question, “Any professional diagnosis of diabetes mellitus?” Upon the receipt of a questionnaire reporting newly diagnosed diabetes, participants were sent a supplemental diabetes questionnaire which collected data on date of diagnosis, symptoms at the time of diagnosis (polyuria, polydypsia, weight loss, visual changes, or coma), blood glucose levels, glycosuria, history of ketoacidosis at the time of diagnosis and hypoglycemic medication. Prior to 1996, participants were considered to have diabetes if they met any of the following: 1) ≥1classic symptom of diabetes (polydypsia, polyuria, weight loss, hunger) and an elevated plasma glucose (fasting glucose of 7.8mmol/l [140mg/dl], nonfasting glucose≥11.1mmol/l [200mg/dl], or 2-hour glucose levels≥11.1mmol/l on glucose tolerance test); or 2) elevated plasma glucose concentrations on at least two different occasions in the absence of symptoms; or 3) hypoglycemic treatment (criteria similar to those of the National Diabetes Data Group [12]). For cases reported after 1996, a lower threshold of fasting plasma glucose (≥7mmol/l [126mg/dl]) was used according to the new guidelines for T2DM diagnosis [13]. Type 2 diabetes mellitus status was determined from an algorithm based symptoms at the time of diagnosis, weight, age at start of insulin and age at diagnosis; exposure status was updated accordingly.

Self-reported diabetes was validated among a subset of 71 participants who completed the supplemental questionnaire with excellent validity [14]. Medical records were reviewed by a physician who diagnosed per the outlined criteria and was blinded to participant responses on the supplemental questionnaire. Fifty-nine cases had complete records, with T2DM confirmed in 57 (97%); 12 of the original 71 cases had incomplete records (i.e. 2 did not have laboratory data and 9 had only 1 set of measurements) but were highly suggestive of T2DM [14].

Outcome Assessment

Data on periodontitis was collected at baseline and each biennial questionnaire using the question, “Have you been professionally diagnosed with periodontitis with bone loss?” Self-reported periodontitis has been validated in this population with high validity against bitewing radiographs among dentists and non-dentists [15, 16]. Among dentists, self-reported periodontitis showed a positive predictive value (PV+) of 0.76 and negative predictive value (PV-) of 0.74 and among non-dentists the PV+ was 0.83 and PV- was 0.69. Hence, the self-reported measure indicates good validity in discriminating periodontitis compared to radiographic bone loss, a well-accepted clinical measure of periodontitis [17].

Data on number of natural remaining teeth was collected at baseline with number of teeth lost over the past two years, in categories, collected on follow-up questionnaires using the question, “How many natural teeth have you lost since January 1 (previous questionnaire cycle)?” Self-reported number correlates well with clinical assessment in the general population [18].

Covariate Ascertainment

Data on several known risk factors for periodontitis/tooth loss and potential confounders were updated on biennial questionnaires, with the exception of height (collected in 1986, 1987 and 1988), profession (1986), race (1986), and alcohol and fruit and vegetable consumption were assessed from FFQs. Updated data were available for smoking status, time since cessation (<1, 1-2, 3-5, 6-9 and 10+ years) and average number of cigarettes/day (0-4, 5-14, 15-24, 25-34, 35-44 and 45+). A comprehensive smoking index (CSI) was calculated [19] using a complex algorithm to account for subjects’ reports of duration of smoking, intensity and time since cessation. Based on prior knowledge, the biologic half-life of smoking’s effect on periodontitis was assumed to be 1.5 years and 6.7 years for tooth loss (half-life parameter for the CSI algorithm) [20]. The index allowed for the parsimonious adjustment of smoking, accounted for interactions between smoking dimensions and eliminated potential collinearity due to highly correlated variables. HPFS is a relatively homogenous population with respect to income, education and health awareness which minimizes confounding by socio-economic status and health related behaviors.

Statistical Analyses

Descriptive analyses for baseline characteristics were conducted for the full cohort and separately by diabetes status at baseline. Multivariable adjusted time-varying Cox models were used to evaluate the association between diabetes status and the incidence of periodontitis and tooth loss separately in subsequent 2-yr follow-up periods. Hazard ratios (HR) are presented with corresponding 95% CIs, with age in months as the underlying time-scale, and should be interpreted as the hazard ratio dependent upon each covariate pattern at each time period. All analyses were left censored and participants contributed person-time from the date of return of the 1986 questionnaire until the earlier of the beginning of the month of return of the first questionnaire reporting periodontitis/tooth loss, death, or January 31, 2006.

We evaluated two multivariable models, Model 1 adjusted for: age (at baseline), race (White/Black/Asian/Other), smoking, body mass index (BMI-kg/m²), fruit and vegetable intake (quintiles), physical activity (quintiles), alcohol consumption (0, 0.1-4.9, 5-14.9, 15-29, 30+ g/day), dental profession (yes/no), and baseline history of stroke, coronary artery bypass surgery or myocardial infarction (yes/no). Model 2 additionally adjusted for number of teeth at baseline (≥25/<25). A priori, based on the literature, we proposed to evaluate potential effect modification of the association between T2DM and periodontitis by age, BMI, smoking, physical activity, fruit and vegetable intake, alcohol consumption, hypertension, dental profession, and number of natural teeth at baseline; significance was evaluated by a likelihood ratio test. We additionally evaluated the association between duration of diabetes and incidence of self-reported periodontitis and tooth loss. To address missing information, data were carried forward ≤1 questionnaire cycle, the median value was substituted for missing data on the CSI variable and the complete case method was utilized thereafter (<5% of observations were missing data on any particular covariate). We examined potential bias due to missing data assumptions (except for data on number of teeth at baseline) by utilizing multiple imputation available in SAS (proc mi and proc mianalyze). We evaluated the influence of misclassification of the outcome on effect estimates using methods proposed by Duffy et al. [21]. Predictive values from prior validation studies conducted by Joshipura and colleagues [15, 22] were calculated separately among dentists and non-dentists. A weighted average of these independent samples was calculated to obtain mean predictive values (PV_m+ and PV_m-). The adjusted HR was estimated by taking the exponential of the log of the uncorrected HR divided by the sum of PV_m+, PV_m- and negative 1. All p-values are two-sided. Analyses were conducted with SAS for UNIX statistical software (version 9.1.3; SAS Institute, Cary, NC).

Results

At baseline, 35,247 dentate men who were free of periodontitis were included in this analysis (Table 1). Over 591,941 person years of follow-up, 3,009 self-reported periodontitis and 10,017 tooth loss events were observed. At baseline, men with T2DM were more likely to report hypertension, were missing more teeth, consumed less alcohol and were less likely to be dentists than men without.

Table 1.

Age-adjusted characteristics of HPFS participants by presence of type 2 diabetes mellitus (T2DM) at baseline 1986

Characteristics	T2DM absent	T2DM present
N (%)	32,962 (94%)	2,285 (6%)
Age (years)	53.4 ± 9.6	54.7 ± 8.9
White %	95	93
Married (%)	91	91
Dentist (%)	57	51
Body mass index (kg/m2)	25.3 ± 3.2	28.1 ± 4.0
Hypertension (%)	20	34
Number of Natural Teeth (%)
25-32	88	85
17-24	10	11
11-16	2	2
1-10	1	1
Smokers (%)
Never	52	46
Former	40	44
Current	8	10
Alcohol (g/day)	11.1 ± 15.1	9.6 ± 14.3
Physical Activity (METs/week)	21.6 ± 29.8	16.0 ± 20.7
Fruit & Vegetable Intake (servings/day)	5.5 ± 2.8	5.3 ± 2.6

Values are means ± SD or percentages and are standardized to the age distribution of the study population, except for age.

kg/m², kilograms per meters squared; METs, metabolic equivalent units

In the age adjusted model, risk of periodontitis was 39% higher in men with T2DM (HR=1.39; 95% CI: 1.22-1.57) compared to men without T2DM (the reference group; Table 2). In multivariable analyses (Model 1), those with T2DM exhibited a 29% increased risk of periodontitis compared to the reference group (HR=1.29; 95% CI: 1.13-1.47). Results were similar when restricted to those with confirmed T2DM (Model 1 HR=1.28, 95% CI:1.12-1.46).

Table 2.

Age and multivariable adjusted hazard ratios (HR) and 95% confidence intervals (CI) for periodontitis and tooth loss by presence of type 2 diabetes mellitus (T2DM)

	T2DM absent			T2DM present
N=35,247	Events	P-Y	HR	Events	P-Y	Age adjusted	Model 1a	Model 2b
Periodontitis	2,726	552,254	1.00	283	39,687	1.39 (1.22-1.57)	1.29 (1.13-1.47)	1.29 (1.13-1.47)
Tooth loss	9,175	545,400	1.00	842	39,081	1.22 (1.14-1.31)	1.10 (1.02-1.18)	1.09 (1.01-1.18)

^aModel 1: Race (White/Black/Asian/Other), smoking (comprehensive smoking index), BMI (kg/m²), fruit and vegetable intake (quintiles), physical activity (quintiles), alcohol consumption (0, 0.1-4.9, 5-14.9, 15-29, 30+ g/day), dental profession (yes/no) and baseline history of stroke, coronary artery bypass surgery or myocardial infarction (yes/no)

^bModel 2: Race (White/Black/Asian/Other), smoking (comprehensive smoking index), BMI (kg/m²), fruit and vegetable intake (quintiles), physical activity (quintiles), alcohol consumption (0, 0.1-4.9, 5-14.9, 15-29, 30+ g/day), dental profession (yes/no), baseline history of stroke, coronary artery bypass surgery or myocardial infarction (yes/no) and number of teeth at baseline (≥25/<25)

T2DM, type 2 diabetes mellitus; HR=hazard ratio; CI=confidence interval; P-Y=person year; BMI=body mass index in units of kilograms per meters squared (kg/m²); g/day=grams/day

In the age adjusted model, T2DM was associated with a 22% increased risk of tooth loss (HR=1.22; 95% CI: 1.14-1.31) compared to those without T2DM (Table 2). When adjusting for factors in multivariable Model 1, the association was substantially attenuated towards the null (HR=1.10; 95% CI: 1.02-1.18) and additional adjustment for number of teeth at baseline did not materially alter the association (Model 2).

The association between T2DM and periodontitis varied significantly by levels of fruit and vegetable intake (Table 3). Among men whose fruit and vegetable intake were below the population median, T2DM was associated with a 49% greater risk of periodontitis compared to the reference group (HR=1.49, 95% CI: 1.23-1.80; p-value for interaction=0.03), while there was no observed association among men with fruit and vegetable intake above the median.

Table 3.

Multivariable association between type 2 diabetes mellitus (T2DM) and periodontitis stratified by key risk factors^a

Stratification Variables	Events	T2DM absent	Events	T2DM present HR (95% CI)	p-value for interaction
Age					0.52
<65	2,220	1.00	231	1.27 (1.10-1.46)
≥65	506	1.00	52	1.39 (1.03-1.87)
BMI (kg/m2)					0.82
<30	2,432	1.00	186	1.28 (1.10-1.49)
≥30	294	1.00	97	1.53 (1.18-2.00)
Smoking					0.16
Never	1,133	1.00	99	1.32 (1.06-1.64)
Former	1,281	1.00	160	1.38 (1.16-1.65)
Current	312	1.00	24	0.81 (0.49-1.35)
Hypertension					0.18
Yes	2,010	1.00	143	1.40 (1.15-1.72)
No	716	1.00	140	1.16 (0.97-1.38)
Profession					0.84
Non-Dentist	1,143	1.00	131	1.36 (1.11-1.65)
Dentist	1,583	1.00	152	1.25 (1.04-1.49)
Alcohol Intake (g/d)					0.51
<median	1,294	1.00	157	1.34 (1.12-1.60)
≥median	1,432	1.00	126	1.24 (1.02-1.51)
Natural teeth at baseline					0.28
<25	424	1.00	56	1.67 (1.20-2.33)
≥25	2,302	1.00	227	1.25 (1.08-1.44)
Physical Activity					0.25
< median	1,454	1.00	191	1.40 (1.19-1.65)
≥median	1,272	1.00	92	1.12 (0.90-1.41)
Fruit & Vegetable Intake					0.03
< median	1,164	1.00	140	1.49 (1.23-1.80)
≥ median	1,562	1.00	143	1.13 (0.94-1.35)

^aAdjusted for age, smoking, alcohol consumption, race, physical activity, fruit and vegetable intake, dental profession and baseline history of stroke, coronary artery bypass surgery or myocardial infarction, except for the stratification variables.

HR, hazard ratio; CI, confidence interval; BMI, body mass index in units of kilograms/meters² (kg/m²); g, grams

In sensitivity analyses, we stratified by smoking status to evaluate the role of residual confounding by smoking and obtained similar results among never smoker (HR=1.32; 95% CI: 1.06-1.64). We additionally adjusted multivariable Model 1 for hypertension status; the association between diabetes and periodontitis and tooth loss were materially unchanged (periodontitis: HR=1.26; 95% CI: 1.11-1.44; tooth loss: HR=1.10; 95% CI: 1.02-1.18). To further evaluate the impact of diet we adjusted multivariable Model 1 for whole grain consumption, however estimates were unaltered (periodontitis: HR=1.29, 95% CI: 1.13-1.47; tooth loss: HR=1.10, 95% CI: 1.02-1.19). Further, we conducted multiple imputation analyses to account for missing data and the results were similar to the primary analyses for periodontitis and tooth loss (periodontitis: HR=1.35; 95% CI: 1.20-1.53, tooth loss: 1.18; 95% CI: 1.10-1.28). We also evaluated the association between duration of T2DM and risk of periodontitis, categorized as ≤10 yrs and >10 yrs compared to those without diabetes. In multivariable adjusted analysis (Model 1), duration of T2DM of ≤10 years was significantly associated with a 26% increased risk of periodontitis compared to those without diabetes (HR=1.26; 95% CI: 1.07-1.49), while >10 yrs was associated with a 33% greater risk (HR=1.33; 95% CI: 1.11-1.61). Duration of diabetes did not exhibit an association with tooth loss (≤10 yrs: HR=1.11; 95% CI: 1.01-1.22; >10 yrs: HR=1.09; 95% CI: 0.97-1.21).

Discussion

In this study with 20 years of follow-up, T2DM was significantly associated with greater risk of self-reported periodontitis. Furthermore, T2DM exhibited an even stronger association with risk of periodontitis among those who consumed few fruits and vegetables.

The variation in the association between T2DM and periodontitis by fruit and vegetable intake are novel and should be replicated. Few studies have investigated the association between fruit and vegetable consumption and periodontitis. Yoshihara et al. [23] reported an inverse association between consumption of dark green and yellow vegetables and incidence of periodontitis over 6 years. Al-Zahrani et al. [24] reported an inverse association between healthy diet (healthy eating index) with odds of periodontitis. The mechanism by which fruit and vegetable intake may mitigate risk of periodontitis among those with diabetes remains unclear. Diet may be directly involved in modulating the inflammatory the process, serve as a proxy for other healthy behaviors, or promote inflammation through diets low in particular fruits and vegetables. Furthermore, the antioxidant benefits of fruits and vegetables may a play a role in ameliorating the oxidant stress induced by hyperglycemia [25].

Few other studies have evaluated the prospective association between diabetes and periodontitis. Overall, the observed associations have been strong, but estimates have varied substantially [6-8, 10]. Three of the [6-8] prospective studies were conducted among the Pima Indians of the Gila River Indian Community in Arizona which have one of the highest known prevalence of T2DM [26]. Thus, the generalizability of the reported effect estimates may be limited due to the population’s unique susceptibility to diabetes and gene-environment interactions. Although it is unlikely the biologic mechanisms underlying this association would differ substantially across populations, gene-environment interactions among sub-groups may exist that limit generalizability since this study consisted primarily of white men.

The strength of the associations presented here are modest compared to those reported in the literature. Previous studies suggest that the association between diabetes and periodontitis may be restricted to poorly controlled diabetes and our data are limited by a lack of information on glycemic control. Taylor et al. [7] reported that glycosylated hemoglobin (HbA_1c) >75mmol/mol (>9.0%) was associated with a significantly greater odds of change in bone score compared to participants without T2DM (OR=11.4; 95% CI:2.5-53.3). In multivariable analysis, Tsai et al. [27] reported HbA_1c≥75mmol/mol (≥9.0%) was significantly associated with a nearly threefold increase in the odds of severe periodontitis (OR=2.90; 95% CI:1.40-6.03). However, the association was not significant among individuals with diabetes under better glycemic control (HbA_1c<75mmol/mol; <9.0%) (OR=1.56; 95% CI: 0.90-2.68). Our population may consist of those with primarily well controlled diabetes, given their presumably good access to medical care and health oriented behaviors potentially attenuating the association. Our binary measure of periodontitis cannot account for severity of disease and may underestimate the true association. While the use of self-reported periodontitis versus clinical measures may result in attenuated estimates, these self-reported measures have been validated against radiographic bone loss with good validity in this population. Moreover, to assess the influence of potential misclassification of the outcome we calculated an adjusted HR. The adjusted estimate for the association between diabetes and periodontitis from Table 2 (Model 1) was HR=1.57, suggesting a 25% attenuation of the unadjusted HR, consistent with the level of attenuation suggested by Joshipura and colleagues [16]. Additionally, we do not have information on periodontal treatment. Greater than 50% of the population are dentists, therefore the level of care is expected to be high. However, periodontal treatment may potentially modify the association and should be explored in further studies.

Hypertension was considered carefully for inclusion in the multivariable models due to a bi-directional relationship with diabetes; results were similar for models with and without hypertension (results not shown).

Diabetes may influence the pathogenesis of periodontitis similar to other macro- and micro-vascular complications of diabetes [28]. Patients with diabetes show evidence of increased pro-inflammatory cytokines within the gingival crevicular fluid and gingival tissues compared to periodontitis patients without diabetes [29]. Hence, diabetes may heighten the inflammatory response launched among affected individuals in the presence of periodontal pathogens, thereby initiating a cycle of tissue destruction and impaired wound healing.

With long-standing hyperglycemia proteins become glycated leading to the formation of advanced glycosylation end-products (AGEs) [29]. AGE formation on proteins (e.g collagen) can lead basement membrane thickening in gingival tissues, impairing the delivery of leukocytes and nutrients into the gingival and periodontal tissues and the movement of metabolic waste of periodontal pathogens out of the tissue; causing decreased wound healing capacity and increased disease severity [29]. Further, AGEs bind to a variety of cell receptors (RAGE) (e.g. endothelial, monocytes and macrophages) [29]. The AGE-RAGE interaction on monocytes leads to activation of transcription factor nuclear factor kappa B (NF-! B) which modifies the monocyte/macrophage phenotype leading to increased production of IL-β and TNF-α [29]. IL-1! plays a key role in tissue destruction by up-regulating host immune response leading to the destruction of periodontal ligaments and alveolar bone resorption [29]. While TNF-α may promote periodontal tissue destruction by inducing bone resorption, endothelial cell apoptosis and by disrupting the migration of periodontal ligament cells [29].

Dyslipidemia may additionally play a role in linking diabetes and periodontitis. Dyslipidemia among individuals with diabetes is generally characterized by altered LDL profile, high concentration of triglycerides (TG), low concentrations of HDL and may be further exacerbated by uncontrolled diabetes, driven by insulin resistance [29]. These lipid abnormalities may be associated with impaired immune response leading to alterations in cellular membrane lipid composition and subsequent impairments of receptor and enzyme channel function [29]. Peripheral blood poly-morphonuclear leukocytes one of the most numerous immune cells present in periodontal lesions, lead to increased production of IL-1β in the presence of high plasma TG [29].

This study has several notable strengths including the prospective design allowing for the appropriate temporal sequence of diabetes prior to periodontitis and tooth loss, while the 20 year follow-up minimizes potential reverse causation. Although residual confounding cannot be ruled out, it is unlikely to explain the observed associations, given minimal attenuation of the estimates between the age and fully adjusted models and minimal confounding by individual key risk factors. Furthermore, while unmeasured confounding is another source of potential bias; this unique population undergoes the collection of biennial questionnaires allow for modeling updated exposure and a comprehensive set of updated confounders in addition to examining variation by key risk factors, including diet. Moreover, the homogenous nature of the cohort minimizes confounding by socio-economic status and health oriented behaviors.

This study provides further support for diabetes as a risk factor for periodontitis. This is among the largest prospective studies to evaluate this association with comprehensive control of confounders. An association between diabetes and periodontitis may potentially affect a large proportion of the population. Moreover, individuals with diabetes may also be at an increased risk of tooth loss due to increased risk of periodontitis. These results hold important public health implications due to the associations between periodontitis and cardiovascular disease and nutritional alterations associated with tooth loss. Greater collaboration between diabetes care providers and dentists could be used to identify at risk patients in both clinical settings.