The inclusion criteria for studies were the following: (i) use of periodontitis as exposure factor and MI as outcome; (ii) observational studies (of all types). There was no language restriction. Case studies, review studies, case series, experimental models, response letters, editorials and duplicated publications were excluded. Duplicate studies were those published by the same study group, with the same inclusion date and population characteristics. In case of duplicate studies, the study with the larger sample size was considered for analysis.

The following databases were used in the literature search, from inception to October 2018: MEDLINE (via PUBMED), EMBASE, Cochrane The Cochrane Controlled Trials Register (CCTR), Latin American Caribbean Health Sciences Literature (LILACS via BIREME), Cumulative Index to Nursing and Allied Health Literature (via EBSCOhost), Scopus and Web of Science (Thomson Reuters). The MEDLINE search strategy was adapted to the other databases.

Two authors independently screened the abstracts and titles of the studies retrieved. Full texts of all potentially relevant articles were then analyzed. In both phases, in case of disagreement between the researchers, a third independent researcher was involved to achieve final consensus. For studies that fulfilled the inclusion criteria, three authors worked in the data extraction process. The agreement between the authors, estimated by the Kappa coefficient, was considered good (k = 0.735).

The following data were extracted from the articles selected: identification (authors and year of publication), general characteristics of participants (including age and gender), sample size, time of observation, definition of MI and periodontitis, magnitude of exposure and effect size.

The association between exposure and outcome was synthesized from a meta-analysis using estimated risk calculations.

Risk of bias was analyzed using the NIH Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies.¹³ This tool allows the assessment of the risk of bias and the quality of the study, and classified them as “good”, “fair” or “bad”. “Good” quality was attributed to studies that adopted at least eight of the established criteria. The biases were assessed by two evaluators and, in case of disagreement, a third evaluator was involved.

The association between periodontitis and MI was analyzed using meta-analysis with random-effects models. The heterogeneity among studies was assessed using the Q-Cochran test and I² statistics. We planned to explore heterogeneity using race as a factor, but the studies did not provide sufficient data. For longitudinal and clinical trials that assessed phase angle in more than one moment, the baseline values were considered for analysis. The authors of the present review were aware that some of issues suitable for sensitivity analysis were only identified during the review process, when the characteristics of each study were identified. At this phase of our review, we performed a sensitivity analysis to assess the robustness of our analyses by including only good-quality studies with good quality. The meta-analysis was performed using the Meta R package (https://CRAN.R-project.org/package=meta).

Of the six articles included in qualitative analysis, the studies of Noguchi et al.,¹⁴ , Lee et al.,¹⁵ Hansen et al.¹⁶ and Holmlund et al.¹⁷ were cohort studies, one retrospective¹⁵ and three prospective studies^{14 , 16 , 17} and the studies by Arbes et al.,¹⁰ and Holmlund et al.¹⁸ were cross-sectional studies. The shortest observation period was 5 years¹⁴ and the largest was 34 years.¹⁷ The sample size of the studies ranged from 3,081¹⁴ to 1,025,340 participants,¹⁵ and their ages ranged from 1¹⁵ to 85¹⁷ years, with two studies without maximum age restriction.¹⁰ Except for the study by Noguchi et al.,¹⁴ who included only men in their sample, the other studies did not show gender restriction. No study used ‘previous diseases’ as inclusion criteria. Data extracted from the selected studies are described in Table 1 . Data regarding the number of outcomes and estimated risks of the six studies included in the qualitative analysis are available in Table 2 .

Table 1

Author, year	population	n	Age (Years)	Gender	Design and observation time	Definition of mi	Definition of periodontitis
Arbes et al., (1999) 10	NHANES III* data of United States population sample	5564	≥40	2,757 men 2,807 women	Cross-sectional study	Self-reported heart attack	Clinical attachment loss ≥ 3mm
HOLMLUNet al., (2006) 17-18	Patients from the Department of Periodontology, Gävle County Hospital	4.254	20 – 70	1,866 men 2,388 women	Cross-sectional study	Self-reported heart attack	Bone loss >2mm
NOGUCHI ET AL., (2014) 14	Employees of a financial company in Japan	3081	36 – 59	Only men	Prospective cohort between 2004 - 2009	Patients report that they were undergoing myocardial infarction treatment	Self-administered questionnaire produced by the authors.
LEE et al., (2015) 15	Random sample (2% of total) of South Korean population covered by National Health Insurance	1.025.340	1-79	513,258 men 512,082 women	Retrospective cohort between 2002 a 2013	Records of the 6th ed. Korean Classification of Diseases (KCD-6): Code not specified	Records of the 6th ed. Korean Classification of Diseases (KCD-6): K052 – K056
HANSEN et al., (2016) 16	Sample of the Danish population from the National Patient Registry	100.694	≥18	57,421 men 43,273 women	Prospective cohort between 1997- 2011	International Classification of Diseases (ICD-10) Records: Codes I21-I22	International Classification of Diseases (ICD-08) Records: Code 523.49 and 523.49 International Classification of Diseases (ICD-10) Records: Code K05.2 - 05.3.
HOLMLUNet al. ( 2017) 17-18	Patients from the Department of Periodontology, Gävle County Hospital	8.999	20-85	3,870 men 5,129 women	Prospective cohort between 1979 - 2013	International Classification of Disease Records: 421 (ICD-8 and ICD-9) and I21 (ICD-10)	Probing depth ≥ 4mm

*NHANES III, Third National Health and Nutrition Examination Survey

Table 2

Author, year	Severity		Control	Outcome (N. cases/N. controls)	HR, ODDS, RR, OR (IC 95%)	Variable
ARBES S.J, (1999)	Percent of sites with periodontal attachment loss of ≥3 mm:	>0-33	No periodontitis	88/5.564	ODDS 1,38 (0,75-2,54)	Percent of sites with periodontal attachment loss ≥3; age; sex; race; smoking status; history of diabetes; history of high blood pressure; body mass indices; poverty index; serum cholesterol.
		>33-67		73/5.564	ODDS 2,28 (1,18-4,39)
		>67-100		55/5.564	ODDS 2,28 (1,18-4,39)
HOLMLUND A, (2006)	Periodontal bone loss		No periodontal bone loss (<2mm)	Not informed	ODDS 2,69 (1,12-6,46)	Age, gender and smoking.
NOGUCHI S, (2014)	With periodontist or without periodontitis		No periodontitis	9/3081	OR 2,26 (0,84-6,02)	Age; BMI; current smoking; hypertension; diabetes; dyslipidemia; family history of heart disease.
LEE J.H, (2015)	With periodontist or without periodontitis		No periodontal disease	8.179/1.025.340	ODDS 0,98 (0,86-1,12)	Sociodemographic/economic status and comorbidities.
HANSEN G.M, (2016)	With periodontist or without periodontitis		No periodontitis	Not informed	IRR 1,16 (1,04- 1,30)	Age; sex; smoking; comorbidities; medication and socioeconomic status.
HOLMLUND A, (2017)	Numbers of surfaces with pocket depth > 4mm		No periodontal disease	570/8.999	IRR 1,13 (0,98-1,30)	Age; sex; education level and smoking.

The association of MI in patients with periodontitis was calculated using data retrieved from four studies^{10 , 14 , 15 , 18} including 1,035,703 subjects. The result of this analysis was – RR: 5.99 (95% CI: 1.17-30.68), but with a high heterogeneity (I² = 100%; p <0.01). Result of the analysis including only the highest-quality articles was lower – RR: 2.62 (95% CI: 1.47-4.70 3.83 ), with a lower heterogeneity also (I² = 85.5%; p < 0.01) ( Figure 2 ).

Figure 2

The quality of the included studies was considered reasonable. All of them described adequately the purposes and the population of the study, and evaluated periodontitis prior to the development of MI. Only in one study (16.67%)¹⁸ the participation rate was less than 50% of the population. No study described the sample calculation in detail, blinded the periodontitis assessment or performed periodontitis measurement more than once over time. The follow-up time described was considered good in four (66.7%) of the studies.^{14 - 17} In only three^{10 , 14 , 18} (50.0%) studies patients were classified according to the degree of periodontitis. Misdiagnosis and/or validation of the diagnosis of acute MI were found in all studies; in contrast, only three^{10 , 17 , 18} (50.0%) studies were considered adequate satisfactory in the definition and validation of the diagnosis of periodontitis. Only one study¹⁸ (16.67%) did not perform statistical adjustment for losses to follow-up, and three (50.0%) studies^{10 , 15 , 17} reported losses to follow-up of less than 20% of the participants.

Publication bias was not evaluated due to the limited number of studies.