Several researchers have examined the role of inflammation

Transcription

1 Asymptotic Dental Score and Prevalent Coronary Heart Disease Sok-Ja Janket, DMD, MPH; Markku Qvarnström, DDS, MS; Jukka H. Meurman, DDS, MD; Alison E. Baird, MD, PhD; Pekka Nuutinen, MD, PhD; Judith A. Jones, DDS, MPH, DScD Background Oral infections have been postulated to produce cytokines that may contribute to the pathogenesis of coronary heart disease (CHD). We hypothesized that by estimating the combined production of inflammatory mediators attributable to several oral pathologies, we might be able to explain CHD with better precision. Methods and Results A total of 256 consecutive Finnish cardiac patients from Kuopio University Hospital with angiographically confirmed CHD and 250 age-, gender-, and residence-matched noncardiac patients (controls) were recruited. All dental factors expected to generate inflammatory mediators, including pericoronitis, dental caries, dentate status, root remnants, and gingivitis, were examined, and an asymptotic dental score (ADS) was developed by logistic regression analyses with an appropriate weighting scheme according to the likelihood ratio. We validated the explanatory ability of ADS by comparing it to that of the Total Dental Index and examining whether the ADS was associated with known predictors of CHD. A model that included ADS, C-reactive protein, HDL, and fibrinogen offered an explanatory ability that equaled or exceeded that of the Framingham heart score (C statistic 0.82 versus 0.80). When ADS was removed from this model, the C-statistic decreased to 0.77, which indicates that the ADS was a significant contributor to the explanatory ability of a logistic model. Conclusions ADS may be useful as a prescreening tool to promote proactive cardiac evaluation among individuals without overt symptoms of CHD. However, additional prospective study is needed to validate the use of an oral health score as a predictor of incident CHD. (Circulation. 2004;109: ) Key Words: oral health heart diseases lipoproteins interleukins fibrinogen Several researchers have examined the role of inflammation in the pathogenesis of atherosclerosis and subsequent coronary heart disease (CHD). 1 3 Inflammatory markers including C-reactive protein (CRP), fibrinogen, and leukocyte counts were found in significantly higher levels among persons with severe gingivitis or periodontitis. 4,5 Potentially supportive data identified antibodies specific to Porphyromonas gingivalis (P. gingivalis) in atherosclerotic plaques of patients with overt periodontitis. 6 8 Because P. gingivalis is a See p 1076 microorganism uniquely indigenous to the oral cavity, these reports suggest that periodontal disease might contribute to atheroma formation in some individuals. In most of these studies, only 1 oral disease, periodontitis, was investigated, although there are many other lesions in the oral cavity that may generate inflammatory mediators such as interleukin 1-B, interleukin 6, interleukin 8, and CRP. Kweider et al 4 observed that severe gingivitis was predictive of elevated inflammatory markers for CHD, such as fibrinogen and leukocytes. Mattila et al 9 incorporated several oral lesions and created a Total Dental Index (TDI) by summing several oral pathologies; the TDI was a significant predictor of CHD in their cohort. This scoring system was based on an arbitrary weighting scheme, and we hypothesized that by using asymptotic weights, we might be able to improve the explanatory ability of the scoring system. From our previous work, 10 we observed a significant increase in many dental disease parameters in CHD patients. Moreover, we postulated that the combination of several oral lesions might provide superior explanatory ability for CHD compared with a single pathological entity. We have used a similar algorithm in predicting good recovery after acute stroke. 11 The purpose of this project was to construct a scoring system useful in explaining CHD using combined oral pathologies that might generate inflammatory mediators. We tested the following hypotheses: (1) The Asymptotic Dental Score (ADS), created by mathematical modeling of oral pathologies, is associated with CHD. (2) Received August 12, 2003; revision received October 31, 2003; accepted November 13, From Boston University Goldman School of Dental Medicine (S.-J.J., J.A.J.), Boston, Mass; Department of Otorhinolaryngology/Oral Surgery (M.Q.) and Department of Cardiothoracic Surgery (P.N.), Kuopio University Hospital, Kuopio, Finland; Institute of Dentistry (J.H.M.), University of Helsinki, Department of Oral and Maxillofacial Diseases, Helsinki University Central Hospital, Helsinki, Finland; National Institute of Neurological Disorders and Stroke (A.E.B.), National Institutes of Health, Bethesda, Md; Harvard School of Public Health (S.-J.J.), Harvard University, Boston, Mass; and VA Center for Health Quality, Outcomes and Economic Research (J.A.J.), Bedford, Mass. Correspondence to Sok-Ja Janket, DMD, MPH, Department of General Dentistry, Boston University Goldman School of Dental Medicine, 100 E. Newton St, Boston, MA sjanket@hsph.harvard.edu 2004 American Heart Association, Inc. Circulation is available at DOI: /01.CIR E 1095

2 1096 Circulation March 9, 2004 Proven hematologic and metabolic predictors of CHD, including CRP, leukocyte count, erythrocyte sedimentation rate, fibrinogen concentration, HDL cholesterol, ratio of total cholesterol to HDL, and serum triglyceride level, are associated with Asymptotic Dental Score (ADS). Methods Ethical and Human Subject Protection Consideration This study was approved by the Joint Ethical Committee of the Kuopio University Hospital and the University of Kuopio, and written informed consent was obtained from all participants. This project adhered to the guidelines set forth by the Declaration of Helsinki and the Belmont Accord to ensure the safety of human research subjects. Participants We recruited 256 consecutive cardiac patients with CHD confirmed by coronary angiography at Kuopio University Hospital. Potential subjects were excluded if they took antibiotics during the previous 30 days or had chronic infection other than dental disease. Also recruited were 250 age- and gender-matched patients from the same catchment area without any evidence of CHD who were admitted to general surgery or otorhinolaryngology at the same hospital. The same exclusion and inclusion criteria were applied to noncardiac patients. Additional exclusion criteria were (1) those who needed emergency coronary bypass surgery or valvular replacement surgery, (2) those whose disease status was so grave a dental examination or dental x-ray could not be taken safely, and (3) those who needed antibiotic prophylaxis before dental probing and examination. Predictor Assessment Panoramic tomograms of the jaws were taken, and signs of dental infection such as periapical radiolucencies, signs of long-standing dental decay or infection manifested either by pericoronitis (defined as an infection/inflammation surrounding the third molars [radiolucent follicle around the retained or erupting third molars with diameter 3 mm]) or numbers of root remnants with soft tissue inflammation (dental hard tissues are usually destroyed by advanced dental caries, leaving only tips of the root), amount of vertical bone loss (measured from cemento-enamel junction in millimeters), calculus deposits, and restorations with overhangs were recorded. A single examiner (MQ) performed all radiographic examinations twice, and agreement of the 2 readings was excellent ( 0.9). The same examiner (MQ) performed clinical dental examinations immediately after panoramic radiography using the World Health Organization format. 12 Dental caries was categorized from 1 to 4 in TABLE 2. Comparison of Explanation Ability of ADS and TDI TABLE 1. General Characteristics of the Cohort CHD Group (n 256) Non-CHD Group (n 250) Mean age SD, y Men, n (%) 163 (64) 159 (64) Women, n (%) 93 (36) 91 (36) Edentulous, n (%) 90 (35.16) 36 (14.40) Dentate, n (%) 166 (64.84) 214 (85.60) No. of teeth, mean SD No. with sound teeth (%) 87 (34.4) 156 (62.4) Current smoker, n (%) 26 (11) 21 (10) Diabetes, n (%) 36 (15) 10 (6) Hypertension, n (%) 117 (50) 47 (24) Total cholesterol 240 mg/dl, n (%) 67 (26) 85 (34) similar fashion as that suggested by Mattila et al, 9 ie, 1 no caries, 2 1 to 3 caries surfaces, 3 4 to 7 carious surfaces or unimaxillary edentulism, and 4 more than 8 carious surfaces or bimaxillary edentulism. Gingivitis was recorded as yes or no. If gingival tissue exhibited overt signs of inflammation, namely, erythema, bleeding, and papillary or generalized swelling, then gingivitis was considered to be present. Periapical lesions, which signify advanced dental caries or periodontal abscess, were categorized in 3 levels: none, 1, and 2 or more. Pericoronitis was recorded as present or absent by clinical examination and radiographic evaluation. Remaining root remnants were categorized in 3 levels: none, 1, and 2 or more. Periodontal disease was measured with the community periodontal index of treatment need (CPITN), and if at least 2 sextants (segments dividing mandible and maxilla into 6) were recorded as having CPITN 3 (signifying that sextant had periodontal pocket depth 3.5 mm), the patient was coded as having periodontal disease. Medical and Clinical Laboratory Examinations A team of cardiologists and cardiac surgeons examined CHD patients according to the Kuopio hospital protocol. A number of blood tests were performed to evaluate serum CRP levels, white blood cell counts, blood fibrinogen level, triglycerides, total cholesterol, HDL, and LDL cholesterol. All blood samples were analyzed immediately. The analyses were performed in batches that included both cases and controls to distribute any potential environmental changes and measurement errors evenly. The erythrocyte sedimentation rate was measured by Westergren s method in glass capillary OR (95% CI) P OR (95% CI) P Logistic model for probability of having CHD (yes/no) 1.79 ( ) ( ) Logistic model for probability of fibrinogen 3 g/l (median) 1.33 ( ) ( ) Logistic model for probability of leukocyte counts ( ) ( ) 0.01 Logistic model for probability of ESR 11 mm/h (median) 1.28 ( ) ( ) 0.48 Logistic model for probability of HDL 1.14 mmol/l (median) 0.86 ( ) ( ) 0.68 Logistic model for probability of total cholesterol/hdl ratio ( ) ( ) 0.33 (median) Logistic model for probability of triglycerides 1.72 mmol/l (median) 1.18 ( ) ( ) 0.81 Logistic model for probability of CRP 3 mg/l 1.39 ( ) ( ) 0.26 *Ordinal scale based on ADS Score by Janket et al. Ordinal scale based on TDI by Mattila et al. 9 Significant at -level of ADS* TDI

3 Janket et al Oral Health Score and Coronary Heart Disease 1097 Figure 1. Receiver operating characteristics curves comparing explanatory ability of ADS alone and ADS plus CRP, HDL, and fibrinogen. tubes, and leukocyte count was measured by Coulter counter. Fibrinogen was measured by the Clauss method. A high-sensitivity immunoturbidometry assay was used to measure CRP with a HITACHI 717 analyzer. Statistical Analyses Using the Statistical Analysis System version 8.2, we evaluated all dental variables postulated to generate inflammatory mediators in a univariate model with CHD (yes/no) as a dependent variable and each dental parameter as a predictor. Sequentially, we added significant variables in the model to evaluate the relationship of variables, examining whether confounding or collinear relationships were present between the variables. Among many variables tested, 5 variables were associated with CHD, including pericoronitis, number of root remnants, dental caries, bimaxillary edentulism, and gingivitis. According to the likelihood ratio, we weighted each variable so that relative importance would be reflected in the prediction score according to the method suggested by Spiegelhalter et al. 13 The final model with appropriate weights might be written as ADS 15 pericoronitis 5 root tips 3 edentulism 4 caries 5 gingivitis, similar to our previous work and the Framingham Heart Score. 11,14 Subsequently, we regressed CHD status on the prediction score and evaluated model fit, and the receiver operating characteristics curve, a global assessment of explanatory ability, was created by plotting sensitivity by (1 specificity), ie, correct identification of CHD versus false-positive identification. Then we proceeded to add other independent CHD risk factors such as CRP, HDL, leukocyte count, and fibrinogen concentration to the model to ascertain whether any of them were confounders or collinear with ADS or with other covariates. Validation The explanatory ability of the ADS was validated by performing logistic regressions with established inflammatory and metabolic markers of CHD as dependent variables and the ADS as a predictor. In addition, we compared the explanatory ability of the ADS to that of the TDI, formulated by Mattila et al. 15 To observe clinically meaningful robust changes, we created an ordinal scale of ADS under the assumption that an increased score would be associated with an increased risk of CHD. To make a fair comparison, we also created a TDI scale with the same assumption. The dependent variables were inflammatory and metabolic markers known to be predictive of CHD, including erythrocyte sedimentation rate, fibrinogen, CRP, leukocyte counts, HDL, triglycerides, and the ratio of total cholesterol to HDL, which was known to be a better predictor of coronary events than any 1 of them alone. 14 The validity of our models was further tested by bootstrapping with 1000 repetitions. Results Demographics and other basic characteristics of the CHD and non-chd groups are listed in Table 1. The groups were well matched with respect to age, gender, and several important factors. However, the CHD group was more likely to be edentulous and had, on average, fewer teeth and fewer sound teeth. Also, the proportions of diabetic subjects and persons with hypertension were higher in the CHD group. The final logistic regression model contained 5 variables; pericoronitis, retained root remnants, edentulism, dental caries, and gingivitis. The Hosmer-Lemeshow test for the final model yielded 0.89, which indicates a good model fit. The area under the receiver operating characteristics curve, equated as the C-statistic, was 0.70, which suggests that this ADS had a slightly lower explanatory ability than the multivariable Framingham Heart score, with C-statistics ranging from 0.73 to The validation results in relation to other independent risk factors of CHD are presented in Table 2. The ADS correctly identified 100% of metabolic and hematologic markers of CHD contained in this data set. In contrast, the TDI scale correctly identified only 3 of 8 markers of CHD. As a univariate predictor, the C-statistic was 0.7 for ADS alone, 0.60 for fibrinogen, 0.60 for triglycerides, 0.63 for

4 1098 Circulation March 9, 2004 HDL, and 0.62 for the ratio of total to HDL cholesterol. In bivariate models with ADS as a main predictor, a model that included ADS and CRP had the best explanatory ability, with a C-statistic of 0.74, whereas C-statistics for other models ranged from 0.72 to When other hematologic or metabolic factors such as CRP, HDL, leukocyte count, and fibrinogen were added 1 by 1, ADS showed signs of confounding by CRP and HDL according to the rule of thumb that a change of more than 10% in parameter estimate can be considered a sign of confounding. This suggested that ADS might be confluent with CRP and HDL. However, ADS remained significant, which indicates ADS is an independent predictor above and beyond the common pathways shared with markers of inflammatory process or lipid metabolism. Leukocyte counts and fibrinogen levels were also confounders of ADS, and leukocyte counts and fibrinogen levels, as expected, were collinear. Because fibrinogen was a much stronger contributor to the explanatory ability, we retained fibrinogen and removed leukocyte counts from the model for the reason of parsimony. This final model consisting of ADS, CRP, HDL, and fibrinogen conferred an 82% explanatory ability, which equaled/ exceeded that of the Framingham Heart Score. 16 This result is presented in Figure 1. When ADS was removed from the best model, the explanatory ability was reduced from 82% to 77%, which suggests that ADS is a significant additional contributor to the explanatory ability of the model containing 3 factors, ie, inflammatory, lipid, and hemostatic factors (Table 3). In the bootstrapping procedure to test the robustness of our results, all the variables remained significant after 1000 repetitions with random selection of variances, which indicates that it is highly unlikely that our results were due to chance. Discussion Our best model (model 6 in Table 3) incorporated the current CHD pathogenic paradigm, which encompasses infection/ inflammation (ADS and CRP), lipid metabolism (HDL), and hemostatic factors (fibrinogen). Although LDL was known to be a strong indicator of atherogenicity, in the present data, HDL was a stronger indicator of CHD. Additionally, LDL was derived from 2 variables, namely, total and HDL cholesterol, and for parsimony, HDL appeared to be a better choice. The ratio of total to HDL cholesterol, considered a better predictor of CHD, 14 was indeed a strong predictor as a single variable (C-statistic 0.62). However, HDL presented a slightly better C-statistic, Thus, our model might be a more parsimonious analog of the Framingham Heart Score in which HDL substituted for total and LDL cholesterol and fibrinogen substituted for smoking, hypertension, and diabetes, because it was reported that fibrinogen was associated with smoking, diabetes, and hypertension. 16 Consequently, our model was able to offer comparable explanatory capability with fewer variables. Some studies reported that periodontal disease might contribute to the generation of inflammatory mediators. 5,17,18 In a meta-analysis of 5 cohort studies, an increased relative risk of CHD due to periodontal disease was reported, but dental disease was considered as a confounder for socioeconomic TABLE 3. Comparison of Explanatory Ability of Different Models P for Parameter Estimate C-Statistic Area Under the Curve* OR Variables (95% CI) Model 1 ADS (linear) 1.17 ( ) Model ADS 1.79 ( ) CRP 7.18 ( ) Model ADS 1.84 ( ) CRP 6.01 ( ) HDL 0.32 ( ) Model ADS 1.83 ( ) CRP 6.50 ( ) HDL 0.50 ( ) Total/HDL ratio 1.28 ( ) Model ADS 1.79 ( ) CRP 3.59 ( ) HDL 0.34 ( ) ESR 1.80 ( ) 0.06 Model 6 ADS 1.71 ( ) CRP 4.57 ( ) HDL 0.37 ( ) Fibrinogen 3.46 ( ) Model 7 CRP 4.94 ( ) HDL 0.40 ( ) Fibrinogen 4.28 ( ) ESR indicates erythrocyte sedimentation rate. *Area under the receiver operating characteristics curve: a global assessment of prediction ability of the model. Best prediction model. ADS removed from the best model. and behavioral risk factors for CHD, presumably because well-conducted epidemiological studies reported null results. 19 A recent meta-analysis 20 of 9 studies also yielded a modest but significant increase in relative risk similar to the result of the previous meta-analysis 21 among individuals with periodontal disease compared with those without. However, the subgroup analyses performed in the latest meta-analysis indicated that there was a significant underestimation (29.7%) of relative risk in epidemiological studies that used selfreported periodontal status. 20 This attenuation due to nondifferential misclassification was remarkably similar to the reported attenuation of 30% by Joshipura et al. 22 In addition, when oral health status was measured by the number of teeth, which is a more precise assessment than a patient s report of past history of periodontitis, a significant association between oral health and risk of stroke was observed in the same cohort. 22

5 Janket et al Oral Health Score and Coronary Heart Disease 1099 Figure 2. Hypothetical diagram of biological pathways linking oral health to systemic health. Although ADS was significantly associated with CHD, we cannot derive any causal inferences. Resolution of the issue of whether oral health status is a contributor or confounder of CHD may be answered by randomized trials. For ethical and financial reasons, a primary prevention trial assigning periodontal disease to examine the relation with future CHD is not feasible. In addition, secondary prevention does not confer an unbiased biological relationship, because individuals who have had prior CHD are at a higher risk. 23 Some studies reported that the intake of some nutrients and foods protective against CHD, such as fruits, vegetables, and fibers, was lower in edentulous patients, and thus it is possible that oral health indirectly affected the risk of CHD via nutrition intake Other studies reported that carbohydrates with high glycemic index were associated with elevated CRP, which suggests that dietary factors may contribute to the inflammatory process. 28 It has been reported that edentulous subjects tend to take in higher levels of carbohydrates, 24 which may possibly increase the level of CRP. After reviewing all of this evidence, we offer an alternative hypothesis regarding socioeconomic factors and oral health in Figure 2. Unfortunately, some of the traditional CHD risk factors were not available in our data set. Nevertheless, our model with ADS, CRP, HDL, and fibrinogen was on a par with the prediction ability of the Framingham model. Because smoking, diabetes, and hypertension were often associated with abnormal fibrinogen, 16 by substituting fibrinogen for them, our model achieved comparable explanatory capability with a more parsimonious model. The present results clearly reaffirm previous comments that dental disease may be partly contributive and definitely predictive. 29 However, further research is needed to validate the ADS in prospective cohort studies. Study Limitations Because our control subjects were selected from hospital patients, selection bias might be a potential problem. However, considering these controls were from the same catchment area where cases arose, it is unlikely that the effect of selection bias affected our results. Because the most severe cases of CHD were excluded, the present study may be a very conservative estimation of the explanatory ability of ADS. As seen in Table 1, it appeared that the CHD group might have modified their lifestyles and that the cases and controls became quite similar in respect to other CHD risk factors such as smoking reduction and lowering their cholesterol levels. This might have helped us to detect a subtle contribution of oral health to the pathogenesis of CHD. 30 Conclusions The ADS, which asymptotically summed 5 oral pathologies that were expected to contribute to the generation of inflammatory mediators, was significantly associated with CHD. The ADS may be useful as a prescreening tool for subjects without overt cardiac symptoms to encourage them to seek early cardiac evaluation. Acknowledgment Dr Meurman is supported by a grant from the Paivikki and Sakari Sohlberg Foundation, Helsinki, Finland.

6 1100 Circulation March 9, 2004 References 1. Beck JD, Offenbacher S, Williams R, et al. Periodontitis: a risk factor for coronary heart disease? Ann Periodontol. 1998;3: Danesh J, Collins R, Peto R. Chronic infections and coronary heart disease: is there a link? Lancet. 1997;350: Ridker PM, Hennekens CH, Buring JE, et al. C-reactive protein and other markers of inflammation in the prediction of cardiovascular disease in women. N Engl J Med. 2000;342: Kweider M, Lowe GD, Murray GD, et al. Dental disease, fibrinogen and white cell count: links with myocardial infarction? Scott Med J. 1993;38: Wu T, Trevisan M, Genco RJ, et al. Periodontal disease and risk of cerebrovascular disease: the first national health and nutrition examination survey and its follow-up study. Arch Intern Med. 2000;160: Choi J, Chung S, Kang H, et al. Establishment of Porphyromonas gingivalis heat-shock-protein specific T-cell line from atherosclerosis patients. J Dent Res. 2002;81: Chiu B. Multiple infections in carotid atherosclerotic plaques. Am Heart J. 1999;138:S534 S Li L, Messas E, Batista EL, et al. Porphyromonas gingivalis infection accelerates the progression of atherosclerosis in a heterozygous apolipoprotein E-deficient murine model. Circulation. 2003;105: Mattila KJ, Nieminen MS, Valtonen VV, et al. Association between dental health and acute myocardial infarction. BMJ. 1989;298: Meurman J, Qvarnström M, Sok-Ja Janket S, et al. Oral health and health behavior in patients referred for open-heart surgery. Oral Surg Oral Med Oral Pathol Oral Radiol Endod. 2003;95: Baird AE, Dambrosia J, Janket S, et al. A three-item scale for the early prediction of stroke recovery. Lancet. 2001;357: Dieudonne B. WHO epidemiological surveys on oral health. Int Dent J. 1990;40: Spiegelhalter D, Crean G, Holden R, et al. Taking a calculated risk: predictive scoring systems in dyspepsia. Scand J Gastroenterol. 1987; 128(suppl): Anderson K, Wilson P, Odell P, et al. An updated coronary risk profile: a statement for health professionals. Circulation. 1991;83: Mattila KJ, Valtonen VV, Nieminen M, et al. Dental infection and the risk of new coronary events: prospective study of patients with documented coronary artery disease. Clin Infect Dis. 1995;20: Wilson PW, D Agostino RB, Levy D, et al. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97: Beck J, Garcia R, Heiss G, et al. Periodontal disease and cardiovascular disease. J Periodontol. 1996;67: Wu T, Trevisan M, Genco RJ, et al. Examination of the relation between periodontal health status and cardiovascular risk factors: serum total and high density lipoprotein cholesterol, C-reactive protein, and plasma fibrinogen. Am J Epidemiol. 2000;151: Danesh J. Coronary heart disease, Helicobacter pylori, dental disease, Chlamydia pneumoniae, and cytomegalovirus: meta-analyses of prospective studies. Am Heart J. 1999;138:S434 S Janket S, Baird A, Chuang S, et al. Meta-analysis of periodontal disease and risk of coronary heart disease and stroke. Oral Surg Oral Med Oral Pathol Oral Radiol Endod. 2003;95: Joshipura KJ, Rimm EB, Douglass CW, et al. Poor oral health and coronary heart disease. J Dent Res. 1996;75: Joshipura KJ, Hung HC, Rimm EB, et al. Periodontal disease, tooth loss, and incidence of ischemic stroke. Stroke. 2003;34: Mehta RH, Bossone E, Eagle KA. Perioperative cardiac risk assessment for noncardiac surgery. Cardiologia. 1999;44: Johansson I, Tidehag P, Lundberg V, et al. Dental status, diet and cardiovascular risk factors in middle-aged people in northern Sweden. Community Dent Oral Epidemiol. 1994;22: Joshipura KJ, Willett WC, Douglass CW. The impact of edentulousness on food and nutrient intake. J Am Dent Assoc. 1996;127: Nowjack-Raymer RE, Sheiham A. Association of edentulism and diet and nutrition in US adults. J Dent Res. 2003;82: Lowe G, Woodward M, Rumley A, et al. Total tooth loss and prevalent cardiovascular disease in men and women: possible roles of citrus fruit consumption, vitamin C, and inflammatory and thrombotic variables. J Clin Epidemiol. 2003;56: Liu S, Manson J, Buring J, et al. Relation between a diet with a high glycemic load and plasma concentrations of high-sensitivity C-reactive protein in middle-aged women. Am J Clin Nutr. 2002;75: Wehrmacher W. Periodontal disease predictive and possibly contributive to acute myocardial infarction. Dent Today. 2001;20: Rothman KJ, Poole C. A strengthening programme for weak associations. Int J Epidemiol. 1988;17: