Tooth loss is independently associated with poor outcomes in stable coronary heart disease

Abstract

Objective

We investigated associations between self-reported tooth loss and cardiovascular outcomes in a global stable coronary heart disease cohort.

Methods

We examined 15,456 patients from 39 countries with stable coronary heart disease (prior myocardial infarction, prior revascularisation or multivessel coronary heart disease) in the STABILITY trial. At baseline, patients reported number of teeth (26–32 (all), 20–25, 15–19, 1–14 and no teeth) and were followed for 3.7 years. Cox regression models adjusted for cardiovascular risk factors and socioeconomic status, determined associations between tooth loss level (26–32 teeth: lowest level; no teeth: highest level) and cardiovascular outcomes.

Results

After adjustment, every increase in tooth loss level was associated with an increased risk of the primary outcome, the composite of cardiovascular death, non-fatal myocardial infarction and non-fatal stroke (hazard ratio 1.06; 95% confidence interval 1.02–1.10), cardiovascular death (1.17; 1.10–1.24), all-cause death (1.16; 1.11–1.22) and non-fatal or fatal stroke (1.14; 1.04–1.24), but not with non-fatal or fatal myocardial infarction (0.99; 0.94–1.05). Having no teeth, compared to 26–32 teeth, entailed a significantly higher risk of the primary outcome (1.27 (1.08, 1.49)), cardiovascular death (1.85 (1.45, 2.37), all-cause death (1.81 (1.50, 2.20)) and stroke (1.67 (1.15, 2.39)).

Conclusions

In this large global cohort of patients with coronary heart disease, self-reported tooth loss predicted adverse cardiovascular outcomes and all-cause death independent of cardiovascular risk factors and socioeconomic status.

Tooth loss, stable coronary heart disease, prevention, risk prediction, Clinical trial registration: www.clinicaltrials.gov, NCT00799903

Introduction

In recent decades, periodontal disease (PD) has emerged as a potential risk factor for coronary heart disease (CHD).^1,2 Similar to CHD, PD is a widespread disease with chronic inflammatory properties, influenced by numerous risk factors, of which poor oral health habits, age, smoking, diabetes and socioeconomic status are the most significant.^3,4 PD is initiated by the formation of bacterial plaques between the tooth and gingiva, inducing a host-mediated inflammatory response that may persist and damage the deeper dental support tissues, eventually resulting in tooth loss after many years of disease.⁵ While tooth loss can also have other causes, caries and PD are the most important, with the latter being predominant in older populations.⁶

As there is no consensus on a uniform diagnostic definition of PD, the assumption of an independent link between PD and CHD is based on results from heterogeneous studies that used different exposure variables and had varying outcome measures.⁷ Moreover, all preceding studies are geographically limited, and in the context of established CHD, evidence of an association between PD and recurrent events is scarce, conflicting, and has been deemed insufficient.^8–10 Self-reported tooth loss, a marker of PD and oral health, has previously been associated with multiple cardiovascular risk factors in patients with chronic CHD,¹¹ but its relationship with outcomes is unknown.

The objective of this report was to assess the association between self-reported tooth loss and long-term cardiovascular outcomes in a global chronic CHD population.

Methods

Study population

The STabilization of Atherosclerotic plaque By Initiation of darapLadIb TherapY (STABILITY) study evaluated the efficacy of darapladib, an oral inhibitor of lipoprotein-associated phospholipase A₂ compared to placebo. STABILITY included 15,828 participants from 39 countries on five continents.¹²

Patients were eligible to participate if they had CHD defined as prior myocardial infarction (MI), prior coronary revascularisation, or multivessel CHD without revascularisation. At least one of the following additional enrichment criteria were also required: age ≥60 years; diabetes mellitus requiring pharmacotherapy; high-density lipoprotein cholesterol <1.03 mmol/l; current or previous smoker defined as five or more cigarettes per day on average; moderate renal dysfunction (estimated glomerular filtration rate ≥30 to <60 ml/min/1.73 m² or urine albumin:creatinine ratio ≥30 mg albumin/g creatinine); or polyvascular disease. Patients with an estimated glomerular filtration rate <30 ml/min/1.73 m² were excluded. More detailed accounts of the main study design and outcomes have previously been published.^12,13

All patients provided written informed consent. The relevant ethics committees in each participating country approved the study.

Data collection

A physical exam, blood samples and medical history were obtained prior to randomisation. All blood sample analyses were performed with standardised methods at central laboratories. The estimated glomerular filtration rate was calculated using the modification of diet in renal disease formula.¹⁴ At baseline, patients completed a lifestyle questionnaire, which included questions on education, psychosocial factors, diet, alcohol consumption, physical activity and number of teeth. A total of 15,456 (97%) participants reported number of teeth according to the following categories: 26–32 (all), 20–25, 15–19, 1–14 and no teeth. In this report, these categories were labelled ‘tooth loss levels’, with 26–32 representing the lowest level and no teeth representing the highest.

Study outcomes

The primary outcome was a composite of the first occurrence of cardiovascular death, non-fatal MI, or non-fatal stroke, collectively termed major adverse cardiovascular events (MACEs). Secondary outcomes included non-fatal or fatal MI, non-fatal or fatal stroke, cardiovascular death and all-cause mortality. The cardiovascular death endpoint included death from an unknown cause, fatal MI, fatal stroke, complications of a cardiac procedure, arrhythmia, congestive heart failure/shock, other vascular cause of death and sudden death. All suspected endpoints were initially documented and reported by STABILITY study investigators and subsequently adjudicated according to prespecified criteria by an independent clinical events committee, blinded with respect to the assigned treatment group (darapladib vs. placebo).¹³

Statistical analyses

Continuous variables are presented as the mean and standard deviations, and discrete variables as frequencies and percentages. P values indicating a trend in baseline factors were generated using analysis of variance or the Spearman correlation test. Cox proportional hazards models were used to determine associations between tooth loss levels and outcomes. Tooth loss levels were analysed as ordinal categories. Both the assumption of linearity and proportional hazards were tested. Linearity on the log hazards scale was verified by comparing log likelihood χ² values between the linear and more flexible models. Three models were used adjusting for an increasing number of prespecified baseline factors with established or possible significance as confounders. Model 1 adjusted for randomised treatment (darapladib or placebo) alone. Model 2 adjusted for model 1, age, systolic blood pressure, diastolic blood pressure, body mass index, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, history of diabetes, prior MI, gender, smoking status and waist–hip ratio. Model 3 adjusted for all factors in model 2 with the addition of estimated glomerular filtration rate, family history of CHD, alcohol consumption, years of education, level of physical activity and World Bank country income level (lower middle, upper middle, or high). Cumulative event rates over time were estimated using the Kaplan–Meier method. A P value less than 0.05 was defined as statistically significant.

Analyses were performed at the Duke Clinical Research Institute in Durham, NC, USA using SAS version 9.2 (SAS Institute, Inc., Cary, NC, USA).

Results

A total of 15,456 patients were included in the study, with a median follow-up time of 3.7 years. The baseline characteristics (Supplementary Table 1) show that patients with higher tooth loss levels were more likely to be older, smokers, female and less physically active; these patients were also more likely to have diabetes, higher blood pressure, higher body mass index, a lower education level and poorer kidney function. A more detailed account of the associations observed between tooth loss and cardiovascular risk factors in the STABILITY population has recently been published.¹¹ There were no differences observed between the two treatment groups for any outcome, irrespective of tooth loss level. For all outcomes, risk factor associations observed after adjustment for randomised treatment alone were attenuated after adjusting for cardiovascular risk factors and socioeconomic status (Table 1). This attenuation was largely an effect of adjustment for the traditional cardiovascular risk factors in model 2, with little additional impact of variables added in model 3.

Table 1.

Hazard ratios^a for every increase in tooth loss level (26–32 (lowest level) to no teeth (highest level)).

Outcome	Nb	Model 1	P value	Model 2	P value	Model 3	P value
MACE	1543	1.16 (1.12, 1.20)	<0.001	1.08 (1.04, 1.12)	<0.001	1.06 (1.02, 1.10)	0.004
Cardiovascular death	705	1.30 (1.23, 1.37)	<0.001	1.18 (1.12, 1.25)	<0.001	1.17 (1.10, 1.24)	<0.001
MI	746	1.07 (1.02, 1.13)	0.008	1.01 (0.95, 1.06)	0.849	0.99 (0.94, 1.05)	0.822
Stroke	301	1.22 (1.13, 1.33)	<0.001	1.13 (1.04, 1.24)	0.005	1.14 (1.04, 1.24)	0.007
All-cause death	1120	1.30 (1.25, 1.36)	<0.001	1.17 (1.12, 1.23)	<0.001	1.16 (1.11, 1.22)	<0.001

a

Estimates are hazard ratios (HR, 95% confidence intervals) derived using Cox proportional hazards models including tooth loss level as an independent ordinal variable. HRs are expressed per one increase in level. Model 1 is adjusted for randomised treatment (darapladib or placebo) only. Model 2 is adjusted for model 1 and age, systolic BP, diastolic BP, body mass index, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, history of diabetes, prior MI, gender, smoking status and waist hip ratio. Model 3 is adjusted for model 2 and estimated glomerular filtration rate, family history of coronary heart disease, alcohol consumption, years of education, level of physical activity and country income level.

b

Number of patients with an event in model 1.

MACE: major adverse cardiac event; BP: blood pressure; MI, myocardial infarction.

Major adverse cardiovascular events

During follow-up, there were 1543 patients with MACEs. After adjusting for cardiovascular risk factors and socioeconomic status, there was a 6% increased risk of MACEs associated with every increase in level of tooth loss (Table 1). Patients in the highest tooth loss level (no teeth) had a 27% higher risk of MACEs compared to those in the lowest (26–32 teeth) after model 3 adjustment (Table 2). The estimates of event rates by tooth loss category showed a continuous separation of event curves that continued throughout follow-up (Figure 1(a)).

Table 2.

Hazard ratios for each tooth loss level relative to 26–32 (lowest level).

	Tooth loss levela
	Lowest				Highest
Outcome	26–32 teeth HR	20–25 teeth HR (95% CI)	15–19 teeth HR (95% CI)	1–14 teeth HR (95% CI)	No teeth HR (95% CI)
MACE	1	1.06 (1.02, 1.10)	1.13 (1.04, 1.22)	1.19 (1.06, 1.35)	1.27 (1.08, 1.49)
Cardiovascular death	1	1.17 (1.10, 1.24)	1.36 (1.21, 1.54)	1.59 (1.32, 1.91)	1.85 (1.45, 2.37)
MI	1	0.99 (0.94, 1.05)	0.99 (0.88, 1.11)	0.98 (0.82, 1.17)	0.97 (0.77, 1.23)
Stroke	1	1.14 (1.04, 1.24)	1.29 (1.07, 1.55)	1.46 (1.11, 1.92)	1.67 (1.15, 2.39)
All-cause death	1	1.16 (1.11, 1.22)	1.35 (1.22, 1.48)	1.56 (1.35, 1.80)	1.81 (1.50, 2.20)

a

Estimates are hazard ratios (HR, 95% confidence intervals (CI)) by tooth loss level derived using Cox proportional hazards models where the effect of tooth loss is estimated per category increase. Estimates are adjusted according to model 3.

MACE: major adverse cardiac event; MI, myocardial infarction.

Figure 1.

Cumulative incidence rates by tooth loss level: (a) Major adverse cardiovascular events (MACEs) (cardiovascular death (CV dth), myocardial infarction (MI), or stroke); (b) cardiovascular death; (c) stroke; and (d) all-cause death from 0 to 1400 days from randomisation, stratified by tooth loss level (26–32, 20–25, 15–19, <15, or no teeth).

Cardiovascular death and all-cause death

Cardiovascular and all-cause death occurred in 705 and 1120 patients, respectively. There was a gradually increased risk of cardiovascular death (17% in model 3) and all-cause death (16% in model 3) for every tooth loss level increase (Table 1). When compared to the lowest tooth loss level, the highest level entailed an 85% and 81% increased risk of cardiovascular and all-cause death, respectively, after adjustment according to model 3 (Table 2). The event curves across tooth loss categories for cardiovascular death showed a consistent separation, with the two lowest levels of tooth loss having similar event rates (Figure 1(b) and 1(d)).

Myocardial infarction

There was a total of 746 patients with fatal or non-fatal MI. When adjusted according to model 1, there was a 7% increased risk of MI for every increase in tooth loss level. However, after further adjustment, no significant association remained (Table 1).

Stroke

Fatal or non-fatal stroke occurred in 301 patients. Every increase in tooth loss level was associated with a 14% (model 3) increased risk of stroke (Table 1). Individuals with the highest level of tooth loss had a 67% higher risk of stroke compared to those in the lowest level, after complete adjustment (Table 2). Estimates of event curves by tooth loss category showed a continuous separation of curves that continued throughout follow-up (Figure 1(c)).

Discussion

In this large contemporary study of patients with stable CHD, self-reported tooth loss was robustly associated with increased rates of the primary outcome (a composite of cardiovascular death, non-fatal MI and non-fatal stroke), cardiovascular death, all-cause death and stroke. The associations were independent of a wide range of cardiovascular risk factors and comorbidities associated with recurrent ischaemic events, death and socioeconomic status. We did not detect an association between the level of tooth loss and risk of MI.

To our knowledge, the relationship between tooth loss and outcomes in CHD patients has not previously been evaluated in a prospective study. Therefore, our results provide generalisability to new patient populations and enhance the global scope of the concept of oral disease-associated cardiovascular risk. While no association was detected for MI, we found that tooth loss entailed a gradually increased risk for all other outcomes, which is consistent with a few previous studies in non-CHD populations.^15–18 For example, the Health Professionals Follow-up Study reported an increased risk of non-fatal and fatal CHD (risk ratio (RR) 1.36; 95% confidence interval (CI) 1.11–1.67) in participants with ten or fewer teeth. The risk was higher (RR 1.79; 95% CI 1.34–2.40) when only fatal events were included,¹⁵ signalling an attenuated effect on total risk by including non-fatal MI, similar to our findings. Furthermore, a Swedish study of 7674 individuals evaluated a number of clinical periodontal measures, and found a dose-dependent association with cardiovascular and all-cause death only for number of teeth, but not with the other PD variables.¹⁶ However, our findings are in contrast to those from a general population sample in Finland, in which the association between tooth loss and CHD mortality was eliminated after multivariable adjustment.¹⁹

The high reliability and detail of our baseline data, paired with careful endpoint ascertainment, provide a good basis for the validity of our results. Nevertheless, the lack of an independent relationship between the level of tooth loss and MI was puzzling and differs from a recent large cohort study in which an association with MI was observed, even after comprehensive adjustment.²⁰ This result could have several explanations. First, when subcategorised, the majority of cardiovascular deaths were attributed to ‘unknown cause’ (21%) and ‘sudden death’ (44%). MI was possibly the underlying cause of many of those deaths;²¹ if so, the ‘true’ MI association may, in fact, be stronger.

Second, while tooth loss is a recognised and validated measure of oral health and PD,^6,22 it can also result from conditions other than PD, such as caries. Therefore, tooth loss may not be a sufficiently specific marker of PD capable of exposing an existing relationship with MI. A previous report only observed an association with MI for tooth loss resulting from infection, but not from other causes – information that was unavailable in our study.²³ Third, this lack of association could be subject to effect modification by another variable. For example, age has been suggested as an effect modifier in the PD–CHD relationship, with certain studies finding a positive association only among younger age groups.^24,25 As a result, the relatively high mean age of the entire STABILITY population could explain why an association was not detected. The observation of a strong relationship between oral health and stroke in contrast to the lack of an MI association was intriguing. While reasons for this are unclear, similar findings have been reported previously,²⁶ therefore rendering our findings less likely to be attributable to spurious chance.

Although associations between tooth loss and several established cardiovascular risk factors have previously been demonstrated in this population,¹¹ it is unlikely that the observed relationship with cardiovascular outcomes was mediated by shared risk factors alone, when considering the multivariable adjustment. Several mechanisms by which periodontal pathogens may impact CHD have been proposed, including detrimental effects on endothelial function, atherosclerosis progression and plaque stability.²⁷ Systemic inflammation generated by PD has been suggested as a possible causative pathway based on observations of increased levels of inflammatory markers in patients with PD,²⁸ a finding also previously demonstrated in this cohort with higher levels of C-reactive protein among those with more tooth loss.¹¹ Moreover, while periodontal therapy has reportedly resulted in reduced systemic inflammation²⁹ and improved endothelial function,³⁰ the long-term effects of PD treatment on clinical outcomes are unknown. A study comparing edentulous (who were considered free of concurrent PD) and dentate individuals with concurrent PD, could not find any difference in CHD incidence, leading the authors to suggest that the elimination of chronic infection due to PD does not reduce the risk of cardiovascular events.³¹ This assumption does not, however, take into account the fact that tooth loss reflecting antecedent PD could represent an accumulation of risk over time, rather than the acute effects of concurrent PD, as suggested by our finding of a gradually increased risk of adverse outcomes related to the degree of tooth loss, with edentulous participants being at the highest risk.

Our study had several limitations. First, although the associations between PD and CHD included several cardiovascular outcomes and were independent of multiple risk factors, the possibility of residual confounding or influence by unmeasured causal factors cannot be ruled out. Second, in addition to the previously discussed potential limitations of the exposure variable, tooth loss may also reflect other non-disease-specific aspects such as historical access to dental care, and temporal and geographical trends in dental extractions, possibly rendering it less specific. Third, tooth loss could lead to an altered diet, including reduced consumption of cardiovascular health-promoting items such as fruit and vegetables; however, previous studies adjusting for dietary variables only demonstrated minor attenuations of risk as a result.^20,32 Fourth, in our study, the prespecified tooth loss levels were arbitrarily assigned; as a result, they may introduce imprecision in the analyses. Finally, there was no information on the duration of tooth loss, incident tooth loss during follow-up, denture use or implants, which could potentially affect the results and constitute a source of misclassification.

In conclusion, self-reported tooth loss was associated with a gradually higher risk of the primary endpoint, cardiovascular death, all-cause death and stroke in this large global population. This finding indicates that mechanisms leading to tooth loss, most importantly PD, may contribute to a worse prognosis in CHD, although causality cannot be established from our results. In addition, our study results suggest that self-reported tooth loss can be used to discriminate the risk of adverse events beyond that of established risk factors and socioeconomic status among patients with stable CHD.

Author contributions

All authors have been involved in the study design, analysis and manuscript revision. All authors read and approved the final manuscript. Ola Vedin is the guarantor who accepts full responsibility for the work and the conduct of the study, had access to the data and controlled the decision to publish.

Acknowledgements

The author(s) would like to thank all participants, study personnel and investigators in the STABILITY study. They also thank Ebba Bergman, Uppsala Clinical Research Centre and Erin Hanley, Duke Clinical Research Institute, for editorial support.

Clinical trial registration: www.clinicaltrials.gov, NCT00799903.

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article:

OV: institutional research grant from GlaxoSmithKline (GSK), during the conduct of the study; lecture and consultancy fees from Fresenius and Novartis outside the submitted work.

EH: institutional research grant from GSK during the conduct of the study; institutional research grants from AstraZeneca (AZ), Amgen, Sanofi; expert committee member for Sanofi, Ariad; lecture fees from Amgen, outside the submitted work.

AB: research grants/investigator’s fees and honoraria for lectures from GSK, during the conduct of the study; research grants/investigator’s fees and honoraria for lectures from AZ, Bristol Myers Squibb (BMS)/Pfizer; research grants/investigator’s fees from Sanofi-Aventis, Boehringer Ingelheim (BI), Novartis, Eisai, Duke Research Institute, outside the submitted work.

SD: institutional research grant from GSK, during the conduct of the study; consultancy fees, lecture fees and grants from BI, Pfizer, Merck, Servier, Berlin-Chemie Menarini, Novartis, AZ and Sanofi Aventis outside the submitted work.

RAH: grants from GSK, during the conduct of the study; research grants and consultancy fees from Johnson & Johnson, Merck, Novartis, The Medicines Company; research grant and other from Regado; stockholder in and consultancy fees from MyoKardia; research grants from AZ, BMS, NHLBI, Portola, Sanofi-Aventis, CSL-Behring; consultancy fees from Vida Health, Vox Media, Apo Pharma, Medtronic, WebMD, Gilead Sciences, Orexigen, Adverse Events, Amgen, Daiichi-Lilly, Janssen; other from Evidint, Scanadu; stockholder in Element Sciences; non-commercial relationship with American Heart Association, outside the submitted work.

WK: consultancy fees from GSK, during the conduct of the study; research grants from Roche Diagnostics, Abbott, Singulex, Beckmann; lecture fees and consultancy fees from Novartis, Amgen, AZ, Servier; lecture fees from Actavis; consultancy fees from BioInvent, diaDexus, Cerenis, The Medicines Company, Genzyme, Pfizer, Merck Sharpe & Dohme, outside the submitted work.

JS: employee of GSK and stock ownership in GSK.

PS: grant from GSK, during the conduct of the study; grants from AZ, Sanofi Aventis, Pfizer, Daiichi Sankyo, speakers bureau fees from AZ, Sanofi Aventis, Pfizer, Takeda, MSD, honoraria from AZ, Sanofi Aventis, Pfizer, Takeda, MSD, consultant fees from AZ, Sanofi Aventis, Janssen-Cilag LTD, Takeda, MSD, outside the submitted work.

AS: nothing relevant to disclose.

RHAS: grants and non-financial support from GSK, during the conduct of the study.

HPS: grant from Zone n W, speakers bureau from Sanofi, honoraria from WCN outside the submitted work.

MV: grants from GSK, during the conduct of the study; personal fees from AZ, and BI, outside the submitted work.

DV: grants, personal fees and non-financial support from GSK, during the conduct of the study; grants and personal fees from Eli Lilly, Novartis, Servier, Johnson & Johnson, Sanofi Aventis, Bristol-Myer Squibb, BI, Roche, Pfizer, Berlin Chemie Menarini; personal fees from Abbott, AZ, outside the submitted work.

LW: institutional research grants, honoraria, consultancy fees, travel support and lecture fees from GSK, during the conduct of the study; institutional research grants, consultancy fees, travel support and lecture fees from AZ, BMS/Pfizer; institutional research grants, consultancy fees, and lecture fees from BI; institutional research grants from Merck & Co, Roche; consultancy fees from Abbott, outside the submitted work.

HDW: research grants and personal fees from GSK, during the conduct of the study; research grants and consultancy fees from Daiichi-Sankyo Pharma Development; research grants and advisory board member for AZ; research grants from Sanofi-Aventis, Eli Lilly, National Institute of Health, Merck Sharp & Dohme, outside the submitted work.

CH: institutional research grant from GSK, during the conduct of the study; institutional research grants from Merck, Roche, BMS; grants, advisory board member and lecture fees from AZ, outside the submitted work.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The STABILITY study and the presented analysis were funded by GlaxoSmithKline.