Article Text

Download PDFPDF

Extended report
Association between a history of periodontitis and the risk of rheumatoid arthritis: a nationwide, population-based, case–control study
  1. Hsin-Hua Chen1,2,3,
  2. Nicole Huang2,
  3. Yi-Ming Chen1,2,
  4. Tzeng-Ji Chen2,3,4,
  5. Pesus Chou2,5,
  6. Ya-Ling Lee2,6,
  7. Yiing-Jenq Chou2,6,
  8. Joung-Liang Lan1,2,3,7,
  9. Kuo-Lung Lai1,
  10. Ching-Heng Lin8,
  11. Der-Yuan Chen1,2,3,7
  1. 1Division of Allergy, Immunology and Rheumatology, Department of Internal Medicine, Taichung Veterans General Hospital, Taichung, Taiwan
  2. 2School of Medicine, National Yang-Ming University, Taipei, Taiwan
  3. 3School of Medicine, Chung-Shan Medical University, Taichung, Taiwan
  4. 4Department of Family Medicine, Taipei Veterans General Hospital, Taipei, Taiwan
  5. 5Institute of Public Health and Community Medicine Research Center, National Yang-Ming University, Taiwan
  6. 6Department of Dentistry, Taipei City Hospital-Heping Fuyou Branch, Taipei, Taiwan
  7. 7Institute of Biomedical Science, Chung-Hsing University, Taichung, Taiwan
  8. 8Department of Medical Research, Taichung Veterans General Hospital, Taichung, Taiwan
  1. Correspondence to Dr Der-Yuan Chen, Division of Allergy, Immunology and Rheumatology, Taichung Veterans General Hospital, No 160, Section 3, Chung-Kang Road, Taichung 40705, Taiwan; dychen{at}


Objective To investigate the association between the risk of rheumatoid arthritis (RA) and a history of periodontitis.

Methods This nationwide, population-based, case–control study used administrative data to identify 13 779 newly diagnosed patients with RA (age ≥16 years) as the study group and 137 790 non-patients with RA matched for age, sex, and initial diagnosis date (index date) as controls. Using conditional logistic regression analysis after adjustment for potential confounders, including geographical region and a history of diabetes and Sjögren's syndrome, ORs with 95% CI were calculated to quantify the association between RA and periodontitis. To evaluate the effects of periodontitis severity and the lag time since the last periodontitis visit on RA development, ORs were calculated for subgroups of patients with periodontitis according to the number of visits, cumulative cost, periodontal surgery and time interval between the last periodontitis-related visit and the index date.

Results An association was found between a history of periodontitis and newly diagnosed RA (OR=1.16; 95% CI 1.13 to 1.21). The strength of this association remained statistically significant after adjustment for potential confounders (OR=1.16; 95% CI 1.12 to 1.20), and after variation of periodontitis definitions. The association was dose- and time-dependent and was strongest when the interval between the last periodontitis-related visit and the index date was <3 months (OR=1.64; 95% CI 1.49 to 1.79).

Conclusions This study demonstrates an association between periodontitis and incident RA. This association is weak and limited to lack of individual smoking status.

  • Rheumatoid Arthritis
  • Epidemiology
  • Infections
View Full Text

Statistics from


Rheumatoid arthritis (RA) is a common and destructive chronic systemic inflammatory disease with a prevalence of 57.7–99.6 per 100 000 and an annual incidence rate of 14.0–16.5 per 100 000 in Taiwan, after adjustment according to the age structure of the WHO world standard population.1 Its exact cause remains unknown; however, many genetic and environmental factors are associated with its development,2 of which only smoking is a well-established risk factor.3–9

More recently, periodontitis has been found to be associated with RA and has become a candidate risk factor.10–16 The prevalence of periodontitis varies widely. Recent reports estimate that about 50% of the population aged ≥30 years in the USA have periodontitis.17 ,18 Periodontal disease is a broader term used to describe chronic destructive inflammation of periodontal tissue, which includes the soft and hard tissues surrounding the teeth. Periodontitis is caused by chronic bacterial infections of the gingiva with subsequent destruction of the periodontal attachment and alveolar bone.19 The relationship between RA and periodontitis is supported by pathological and immunological data.20 ,21 Porphyromonas gingivalis (P gingivalis), a major pathogenic bacterium of periodontitis, is the sole micro-organism known to express peptidyl arginine deiminase,22 which leads to citrullination, an irreversible, post-translational conversion of arginine to citrulline.15 ,23 It is hypothesised that with the accumulation of citrullinated peptides, immune tolerance to endogenous citrullinated antigens in a genetically susceptible (shared epitope positive) individual may be broken, followed by the development of anti-cyclic citrullinated peptide antibodies, which are implicated in RA development.15 ,24

Multiple epidemiological studies have reported a higher prevalence of periodontitis in patients with RA than in non-RA controls or the general population.14 ,16 ,25–27 Mercado et al reported that a greater number of patients with RA have missing teeth because of periodontitis compared with non-RA controls.28 Compared with non-RA controls, patients with RA have increased gingival bleeding and plaque accumulation;21 ,25 in contrast, Mercado et al reported no such association.28 A recent large-scale prospective study also reported no evidence of increased RA risk in patients with a history of periodontal surgery and/or tooth loss.29 Furthermore, no association was identified in two cross-sectional studies and one retrospective study.30–32 These conflicting data may be attributed to selection or recall bias and variations in sample size, RA and periodontitis classification criteria, patients’ race, and study design. No population-based, case–control study using administrative data has investigated RA risk on the basis of periodontitis history.

The Taiwanese National Health Insurance Research Database (NHIRD) has recently facilitated the release of claims data for population-based longitudinal studies. This study used the NHIRD to estimate the OR for RA development on the basis of periodontitis history.


Study design

This was a nationwide, population-based, case–control study that used claims data.

Data source

The National Health Insurance (NHI) Programme was implemented on 1 March 1995, and covers >98% population in Taiwan. It includes data on ambulatory care, inpatient services, dental services, traditional Chinese medical services and prescription drugs. The NHIRD is managed by the National Health Research Institute and comprises comprehensive NHI-related administrative and claims data for research purposes. However, the NHIRD does not include some personal information, such as smoking and alcohol use. Multiple NHIRD datasets were used for this study: NHI catastrophic illness files, 1996–2006 ambulatory and inpatient claims files and enrolment files.

The NHI registry system for catastrophic illnesses tracks patients with major or catastrophic illnesses, including some autoimmune diseases such as RA and Sjögren's syndrome (SS). Patients with a catastrophic illness certificate are exempt from co-payment. The Bureau of NHI (BNHI) performs a routine validation of diagnoses by at least two specialists who carefully review original medical records, laboratory data, imaging findings and pathological findings of all patients who apply for catastrophic illness registration. The BNHI issues catastrophic illness certificates only to those who meet the classification criteria for major illness. NHI catastrophic illness files include ambulatory and inpatient claims of beneficiaries and are distributed as a package. The 1996–2006 ambulatory and inpatient files provide information on diagnosis, prescription, date of visit/admission and medical expenses. Enrolment files provide demographic and enrolment information. In 2000, the NHIRD also established a representative database from the entire set of enrollees by randomly selecting 1 000 000 subjects in 2000. All enrolment and utilisation information associated with this random sample are available.

Although laboratory and radiographic data are not available in the database, the BNHI routinely audits the accuracy of diagnoses by randomly sampling patient charts. This BNHI audit has enhanced coding accuracy.33 This study was approved by the ethics committee for clinical research at Taichung Veterans General Hospital.

Study samples

All of the cases were defined as newly diagnosed patients with RA (International Classification of Diseases, 9th revision, clinical modification (ICD9-CM) code 714.0) with a certificate of catastrophic illness between 2001 and 2006. During the years 1996–2006, the American College of Rheumatology classification criteria for RA (1987) were used for RA diagnosis.34 A catastrophic illness certificate for RA was approved after at least two qualified rheumatologists validated the diagnosis of RA by thorough review of medical records, laboratory data and imaging findings. Patients diagnosed before 1 January 2001 were excluded. The index date for the study group was identified as the date of the first ambulatory care visit that resulted in RA diagnosis between 2001 and 2006. Because periodontitis occurs less commonly in children and adolescents, patients aged <16 years were excluded.

The controls were selected from the random sample of one million individuals. Individuals with any RA diagnosis recorded between 1996 and 2006 were excluded. The remaining individuals from the registry of beneficiaries (10 for every case), who had made at least one ambulatory visit, were then randomly extracted as controls, which matched the study group for age at first date of RA diagnosis (16–25, 26–35, 36–45, 46–55, 56–65, >65 years), sex and year of index date (index year). The date of the first ambulatory visit during the matched index year was selected as the index date for the controls.


Periodontitis exposure

In Taiwan, BNHI encourages people to receive regular dental check-ups. Dentists may perform scaling for these people who have subclinical periodontitis with a concurrent coding of periodontitis (ICD9-CM Codes 523.3–5) no more than twice per year. In this study, only patients who had at least one ambulatory visit before the index date with a diagnosis of periodontitis (ICD9-CM codes 523.3–5) and concurrently received antibiotic therapy, or periodontal treatment other than scaling, or scaling more than twice per year by certified dentists, were identified as patients with a history of periodontitis. Different definitions of periodontal disease and reference groups were also used to test whether the results remained robust.

Proxy measures of periodontitis severity

Proxy measures of periodontitis severity included the cumulative number of periodontitis-related visits (ie, 1–2, 3–4, 5–7 and ≥8), the cumulative cost of periodontitis-related visits according to the 25th, 50th and 75th centiles, and the receipt of periodontal surgery before the index date. We assumed that these measures were positively correlated with periodontitis severity. The cumulative cost of the periodontitis-related visits was calculated by summing all expenses of outpatient visits with ICD9-CM codes 523.3–5 before the index date. The cumulative costs were converted from new Taiwan dollars (TW$) to US dollars (US$) using a conversion rate of 30 TW$ to 1 US$. Periodontal surgery included subgingival curettage, periodontal flap operation, gingivectomy and gingivoplasty.


Potential confounders included geographical region and pre-existing diseases, including SS and treated type 1 and type 2 diabetes mellitus (DM). Geographical region was selected because the prevalence of some environmental risk factors for RA may differ among regions. SS was selected because it was found to increase wrist and periodontal bone destruction risk in patients with RA,35 and may be associated with periodontitis.36 In Taiwan, the classification criteria proposed in 1986 and the revised classification criteria by the American–European Consensus Group in 2002 were used as criteria for certifying SS as a catastrophic illness before and after 2002, respectively.37 In this study, only those patients diagnosed with ICD9-CM code 710.2 and holding a catastrophic illness certificate for SS were identified as patients with SS. Because RA and periodontitis share similar genetic backgrounds, the major systemic risk factor for periodontitis, DM,38 ,39 was also selected as a potential confounder. Patients who had at least one ambulatory visit with ICD9-CM codes 250.x and a concurrent prescription of antidiabetic drugs were identified as patients with DM.

Statistical analysis

We compared baseline characteristics between cases and controls using a McNemar test for categorical variables. Conditional logistic regression was used to examine the influence of periodontitis and other potential confounders on the risk of RA development, as shown by ORs with 95% CI. A two-tailed p value of <0.05 was considered statistically significant. All statistical analyses were performed using SPSS V.18.0 for Windows (SPSS, Inc, Chicago, Illinois, USA).


We identified 13 779 newly diagnosed patients with RA (aged ≥16 years) with a certificate of catastrophic illness as cases, and 137 790 matched non-RA patients from the 1 000 000-member representative cohort as controls (table 1). The cumulative incidence of RA between 2001 and 2006 was 0.64%. Table 1 shows the baseline demographic data of the two groups. Table 2 shows a detailed comparison of the history of periodontitis and other pre-existing diseases between the two groups. The female to male ratio was 3.4 : 1. The mean age (±SD) was 52.6 (±14.4) years for cases and 52.4 (±15.4) years for controls. The distributions of geographical regions and histories of pre-existing diseases were different between the two groups. An association between a history of periodontitis and newly diagnosed RA (OR=1.16; 95% CI 1.13 to 1.21) was observed. The strength of the association between periodontitis and RA risk (OR=1.16; 95% CI 1.12 to 1.20) was mildly attenuated after adjustment for measured potential confounders (table 3). The area under the curve of the regression model was 0.53 (95% CI 0.52 to 0.53; p<0.001). The magnitude of this association did not change according to different periodontitis definitions and reference groups (table 4). Residence in the eastern region and SS were also associated with an increased RA risk. Patients with DM requiring antidiabetic treatment had decreased RA risk (table 3). There were no obvious differences in these associations between sexes (table 3).

Table 1

Demographic data of patients with RA and non-RA controls

Table 2

Comparison of pre-existing diseases before index date between patients with RA and non-RA controls

Table 3

Adjusted OR with 95% CI for the association between variables and rheumatoid arthritis

Table 4

Sensitivity analyses comparing adjusted OR (aOR) with 95% CI for rheumatoid arthritis on the basis of different periodontitis exposure definitions and reference groups

Table 5 shows that RA risk was strongest when the interval between the index date and the last periodontitis-related visit date was <3 months (OR=1.64; 95% CI 1.49 to 1.79), followed by an interval of 3–6 months (OR=1.25; 95% CI 1.13 to 1.38). RA risk was also higher in patients with more periodontitis-related visits, those who encountered a higher cumulative cost of periodontitis-related visits and those with a history of periodontal surgery.

Table 5

Adjusted OR with 95% CI for associations between rheumatoid arthritis and history of periodontitis based on sex of the patient


To our knowledge, this case–control study is the first to use nationwide, population-based longitudinal administrative data to examine the strength of the association between exposure to periodontitis and RA risk. One important finding was a significant association between a history of periodontitis and subsequent RA development after adjustment for measured confounders. A dose–response pattern was found in the association between periodontitis severity and RA development. These findings may be explained by the potential role of P gingivalis in the development of autoimmunity associated with RA.13 ,15

In this study, we made two assumptions. First, previous periodontitis exposure, if it existed in cases or controls, was most severe in patients who met the criteria of periodontitis exposure, followed by those who had undergone scaling ≤2 times a year and those who had no previous dental visits that resulted in a diagnosis of periodontitis. Second, in patients with a history of periodontitis, periodontitis severity was correlated with the cumulative number of periodontitis-related visits, cumulative cost of these visits and receipt of periodontal surgery before the index date. Our study results showed that patients with a history of periodontitis exposure, although having received treatment, had a slightly higher risk of RA development than patients without a history of periodontitis. This raises a concern that treatment for periodontitis may theoratically reduce RA risk. However, to date, there is no evidence showing that identification and aggressive treatment of periodontitis can prevent RA development in genetically susceptible individuals. Our data also showed no difference in the risk of incident RA according to a history of antibiotic treatment for periodontitis. Therefore, we hypothesise that local and/or systemic treatment for periodontitis may not halt the process of autoimmunity necessary for RA onset, perhaps because these treatments cannot eliminate the accumulated citrullinated proteins in tissues distant from periodontium (such as joints).40

Although the coding of periodontitis may introduce a misclassification bias, analysis showed that the results were not sensitive to the varying definitions of periodontitis. Moreover, considering that it is impossible to calculate the exact misclassification rate and given the assumption that the misclassification rate for periodontitis was the same for each group in our study, the direction of such non-differential misclassification bias was always towards the null.41

In contrast to the findings of our study, a recent prospective study by Arkema et al29 found no association between a history of periodontal surgery and tooth loss (severe periodontal disease) and RA development risk in a large cohort of American women. In Arkema's study, possible enrolment of patients with mild periodontal disease in the comparison cohort may have underestimated the risk for RA development in patients with severe periodontal disease. Additionally, differences in race and sex distribution might also have caused this discrepancy.

Another significant outcome of our study was that the magnitude of association between periodontitis and RA peaked when the lag time of the last periodontitis-related visit was <3 months. However, we cannot rule out the possibility that RA might have developed before the last periodontitis-related visit when the lag time was <3 months. This is because the 1987 American College of Rheumatology classification for RA suggested that the diagnosis of RA could be made when the symptoms persisted for more than 3 months. Therefore, another explanation of this finding is that early RA could be a risk factor for periodontitis.

We also identified other significant risk factors for RA in this study. The association between geographical regions and risk of RA detected in our study may be explained by the geographical variations in the prevalence of some possible environmental risk factors for RA, such as smoking42 and human parvovirus B19,43 in Taiwan.

An unexpected finding was that DM was a protective factor for incident RA. Tentolouris et al tested the hypothesis that patients with pre-existing DM might have a lower RA prevalence than the general population because of immunosuppression associated with DM.44 Although their study results did not prove their hypothesis, they suggested that it may easily reach statistical significance when the number of patients with type 2 DM increased to 1500.44 Clearly, additional cohort studies are warranted to validate this hypothesis.

The marked association between SS and RA development may imply the inclusion of the rheumatoid factor in the classification criteria for SS and/or the initial presentation of sicca symptoms in some patients with RA. Additionally, our data showed that a higher proportion of patients with SS had history of periodontitis compared with non-SS patients (51.5% vs 35.4%, p=0.006). However, the evidence as to whether patients with SS are at increased risk for periodontitis is conflicting.45 ,46

This study had some limitations. First, the lack of information on smoking status in individual patients and controls was a major limitation. However, if the association between periodontitis and RA risk had been introduced by unmeasured smoking status, and given the marked higher prevalence of tobacco use in men than in women (approximately 9 : 1) in Taiwan,47 the magnitude of the association in men should be larger than that in women if the prevalence ratio of smoking between men and women with RA was similar to that between men and women in the general population. However, the absence of sex differences in the association between a history of periodontitis and RA risk does not exclude the possibility that this association was introduced by unmeasured smoking status.

Second, the accuracy of diagnoses based on administrative data is another issue of concern. Although the BNHI routinely samples patient charts randomly to cross-check the quality of claims from all medical institutions, bias due to miscoding or misclassification cannot be completely avoided. However, the accuracy and validity of RA diagnosis is of less concern because the BNHI selected at least two experienced and qualified rheumatologists to validate RA diagnosis by reviewing patients’ medical charts, laboratory data and x-ray images before issuing a catastrophic illness certificate. In addition, the non-differential misclassification bias related to periodontitis diagnosis in both groups could only underestimate the magnitude of the association between a history of periodontitis and RA risk.

Third, a case–control study is limited by its weaker power of causal inference compared with a cohort study.

Finally, an administrative database cannot provide laboratory or clinical data to test the association between periodontitis and anti-cyclic citrullinated peptide antibody levels; furthermore, such a database lacks the clinical information required for assessing periodontitis severity. However, it is reasonable to conclude that periodontitis severity is related to a history of periodontal surgery, more periodontitis-related visits, and higher costs of medical care. Using these data provided us with a method of estimating a dose–response relationship between a history of periodontitis and RA risk.


This large, non-selective, nationwide, population-based study showed an association between a history of periodontitis and RA risk. The magnitude of this association mildly increased when the lag time of the last periodontitis-related visit was shorter or periodontitis was more severe. Although the association between a history of periodontitis and RA risk was weak, its biological plausibility warrants further clinical and basic studies to elucidate the role of periodontitis in RA initiation and aggravation.


We thank the Biostatistics Task Force of Taichung Veterans General Hospital, Taichung, Taiwan, ROC, for assistance with statistical analysis. We thank the members of the Bureau of National Health Insurance, Department of Health and the National Health Research Institutes for providing and managing, respectively, the National Health Insurance Research Database.


View Abstract


  • Contributors H-HC had full access to all the data in the study and takes responsibility for its integrity and the accuracy of the data analysis. Study concept and design: H-HC, D-YC, NH, PC, J-LL, Y-MC, Y-LL. Acquisition of data: H-HC, T-JC. Analysis and interpretation of data: H-HC, D-YC, NH, PC, Y-LL. Drafting of the manuscript: D-YC, Y-MC, K-LL. Critical revision of the manuscript for important intellectual content: D-YC, NH, PC, J-LL, Y-MC, Y-LL, K-LL, C-HL. Statistical analysis: H-HC, NH, C-HL.

  • Funding This study was supported by grant TCVGH-1013803B from Taichung Veterans General Hospital, Taiwan.

  • Competing interests None.

  • Ethics approval Ethics committee of clinical research at Taichung Veterans General Hospital.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.