This systematic review was recorded in Prospective Register of Systematic Review (PROSPERO-CRD42020190445) and conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (28).

The criteria were defined according to the PECO strategy considering observational studies with adult humans (Participants) with periodontitis (Exposure) and without periodontitis (Comparison) and its association with the presence of anxiety (Outcome). The focused question of this review is this: “Is chronic periodontitis associated with anxiety in adults?”

The inclusion criteria considered was observational studies with individuals with chronic periodontitis and individuals with no history of the disease in which data from anxiety levels was evaluated. Studies including adults with neurological or cognitive disorders were excluded. Case reports, reviews, descriptive studies, opinion articles, technical articles, animal studies, and in vitro studies were also excluded.

The following online databases were accessed: PubMed, Lilacs, Scopus, Web of Science, Cochrane, and Open Gray, from October 2020 to December 2020. No restrictions were applied to searches regarding publication date or language.

The search strategies were adapted according to each database (Supplementary Table 1), in which a research alert was created to notify new studies according to the described search strategy.

Relevant citations were saved in a reference manager (EndNote, version X9, Thomson Reuters). First, duplicate results were removed. Then, titles and abstracts were analyzed according to the inclusion and exclusion criteria. Finally, the remaining articles were evaluated based on the full-text reading. Additional citations were sought from the reference list of all previously selected articles. The selection process was conducted independently by two examiners (W.A.B.A and D.S.M.) and verified by a third examiner (R.R.L.) in cases of disagreement.

The eligible articles were extracted and tabulated by two examiners (W.A.B.A. and D.S.M.) and verified by a third examiner (Y.G.S.N.). The following data were recorded: author/year, type of study, participants (sample source, sample size, age, and characteristics of groups), clinical parameters (periodontal evaluation, anxiety evaluation), statistical analysis, and results. In the absence of information about the studies, the authors were contacted by e-mail for up to 4 consecutive weeks.

Two examiners (W.A.B.A. and D.R.F.) independently assessed the methodological quality and risk of bias by the Newcastle-Ottawa protocol for case-control studies (29) and a protocol adapted for cross-sectional studies (30). A third examiner was consulted (R.O.F.) in case of doubts about the evaluation in any criterion.

The protocol for case-control studies consists of domains that analyze the representativeness of cases and definition of controls, group comparability, determination of exposure, and non-response rate. For cross-sectional studies, the protocol covers the representativeness and size of the samples, non-responders, comparability, evaluation of the result, and statistical test. The examiners standardized the checklist to provide valid information and the viability of the methods. Thus, an asterisk (^*) was applied according to the agreement of the questions in each domain, and a dash (-) was used when the topic analysis was not applicable. The criteria used for the quality assessment are available in Supplementary Tables 2, 3.

The selected studies were grouped to analyze the level of evidence of the outcome using the “Grading of recommendations, assessment, development and evaluation” (GRADE) tool, as the narrative evidence profile, which was evaluated using criteria such as study design, risk of bias, inconsistency, indirectness and imprecision (31). As all studies are observational, the level of evidence starts low and, it can be reduced during the analysis if there is a serious or very serious problem, or it can be increased depending on the dose-effect, confounding factors, and the magnitude of effect.

The total number of studies found in the databases was 6,380, based on the search strategy defined by the PECO. From these, 1,724 duplicates were excluded; and 4,621 were removed by title and abstract analysis. Therefore, 35 studies were selected for full-text reading. The methodological characteristics of the studies were observed as study design, considering the inclusion and exclusion criteria, and 24 studies were excluded due to the presence of control group with smokers (32), absence of periodontal measures (33, 34), and lack of a control group (35–55). Therefore, after full-text reading, 11 studies were selected for qualitative analysis (10, 56–65). Figure 1 describes the flowchart of the studies selection according to the PRISMA protocol. No quantitative analysis was performed since the nature and heterogeneity of the studies was high, not making merging of data possible.

All selected studies were observational—six were case-control (10, 56, 59–61, 63), and five were cross-sectional design (57, 58, 62, 64, 65). Studies were performed in several countries, including Brazil (10, 56, 57, 59), Italy (62), Austria (58), Jordan (60), China (61), Israel (63), India (64), and Iran (65). The number of participants involved in each study ranged from 46 to 300 patients. The age of the participants ranged from 14 to 71 years old.

Included studies present a control group without periodontitis and a case group with diagnosed periodontitis through periodontal measurements of several parameters such as clinical attachment level (CAL), plaque index (PI), probing pocket depth (PPD), bleeding on probing (BOP), plaque control record (PCR), gingival index (GI), probing depth (PD), dental mobility (DM), furcations (F), calculus index (CI), papillary bleeding index (PBI) and approximal plaque index (API).

For studies that did not provide the odds ratio (OR) data with a 95% confidence interval (CI), we performed the calculation using the GraphPad Prism 7.0 software (GraphPad Software Inc., La Jolla, CA, USA).

In the study by Vettore et al. (56), two calibrated examiners evaluated CAL, PPD, GI, and PI at six sites per tooth (mesiobuccal, buccal, distobuccal, distolingual, lingual, and mesiolingual) at all teeth excluding third molars. Participants were allocated into three groups according to their PPD levels: control (<4 sites with PPD = 4.0 mm), test group 1 (at least four sites with PPD ≥ 4 mm and ≤ 6 mm), and test group 2 (at least four areas with PPD > 6 mm). Anxiety assessment was performed using the State-Trait Anxiety Inventory (STAI) method. This is one of the most used methods to assess anxiety levels consisting of two self-report subscales that determine how the subject generally feels (T-trace) and how he feels at a given moment (S-state) (66). According to the Non-parametric Spearman linear correlation, moderate CAL frequencies (p < 0.05), moderate PPD frequencies (p < 0.05), and PPD frequencies >4 mm (p < 0.05) were found to be significantly associated with high scores of anxiety trait.

Solis et al. (57) defined study cases as subjects with two or more interproximal sites from different teeth with a CAL of ≥ 6 mm and at least one additional site with a pocket depth (PD) of ≥ 5 mm. Subjects that did not meet this criterion were allocated to the control group. After defining the groups, a calibrated examiner evaluated CAL, PPD, GI, and PI, and the subjects answered the STAI questionnaire. There were no significant differences between groups (p = 0.57; Mann–Whitney test). Thus, the findings of this study do not confirm the hypothesis of an association between periodontitis and anxiety.

In the study of Laforgia et al. (62), an examiner evaluated CAL, PPD, GI, DM, F, and PCR. Anxiety was assessed using the Symptom Check List (SCL-90), which consists of a questionnaire composed of 90 symptoms and nine symptom dimensions: somatization, obsessive-compulsive disorder, interpersonal sensitivity, phobic anxiety, paranoid ideation, psychoticism, hostility, depression, and anxiety (67). The percentage of anxiety was higher in the group with periodontitis (31.48%) than the control (20.37%), implying that the hypothesis of a correlation between periodontal and psychological variables can be confirmed. Thus, to support that information, the linear regression analysis showed a positive correlation between periodontal conditions and SCL-90 scores.

Saletu et al. (58) diagnosed periodontitis based on clinically and radiographically evident loss of attachment. Three groups were distinguished considering the severity of periodontitis: slight periodontitis (CAL of 1–2 mm and/or a bone loss of 10–30%), moderate periodontitis (CAL of up to 4 mm, and/or a bone loss of 30–50%), and severe periodontitis (CAL of ≥ 5 mm and/or a bone loss of >50%). Participants in the control group had no history of the disease and showed no bone loss and no probing depth >4 mm. Subjects with gingivitis were allocated to the control group. Anxiety assessment was performed using the Self-Rating Anxiety Scale (SAS), which consists of 20 items that include questions about psychological and physical symptoms related to anxiety from positive and negative perspectives (68, 69). Responses range from 1 (none/little) to 4 (all the time), and scores that are equal to or exceed a specified cutoff point are an indication of the likely presence of anxiety (69). This study showed that anxiety scores were higher in subjects with periodontitis (Multifactorial analysis of variance; Case: 31.1 ± 10.1; Control: 25.3 ± 4.0; p = 0.018) in addition to a positive correlation of anxiety scores with the severity of periodontitis and CAL levels, under the control of covariates such as age, smoking, and API (Partial correlation; p ≤ 0.05).

In the publication of Vettore et al. (59), participants were divided into three groups based on PPD levels: control (<4 sites with PPD ≤ 4.0 mm; n =20), test group 1 (at least four sites with PPD ≥ 4.0 and ≤ 6.0 mm; n =26), and test group 2 (at least four areas with PPD > 6 mm; n = 20). Anxiety was assessed using the STAI method. Two calibrated examiners assessed CAL, PPD, PI, Gingival Index (GI) at six sites per tooth (mesiobuccal, buccal, distobuccal, distolingual, lingual, and mesiolingual) at all teeth, except third molars. Non-parametric Spearman linear correlation showed a significant correlation between anxiety trait and PPD (p < 0.05) and CAL (p < 0.05).

Castro et al. (10) diagnosed periodontitis considering a CAL ≥ 4 mm and BOP in at least 10 teeth and probing pocket depth PPD ≥ 6 mm in at least five teeth. A single examiner assessed CAL, PPD, and BOP in both study groups. Anxiety was evaluated using the STAI and Beck Anxiety Inventory (BAI) methods. The latter consists of a self-report scale consisting of 21 items or descriptive statements of anxiety symptoms rated by the subject on a four-point scale. The total score allows a classification of anxiety into levels of intensity (minimum, moderate, and severe) (70, 71). The findings showed that there was no significant association between periodontitis and anxiety (STAI-T: Odds Ratio (OR) = 1.003, Confidence Interval (CI) 0.95–1.06, p = 0.912; BAI: OR = 0.946, CI 0.86–1.04, p = 0.257) in the multivariate analysis.

In the study by Ababneh et al. (60), participants were divided into three groups: control, aggressive periodontitis (AP), and chronic periodontitis (CP). The diagnosis of AP considered a CAL > 2 mm in at least two teeth, “arch-shaped” bone loss, and age <45 years. The CP group was composed of subjects with the same characteristics but older than 45 years old. Therefore, we considered for this review only the CP group to maintain the segmentation of the senior periodontitis classification, on which the studies in this review are based. Anxiety was assessed using the Hospital Anxiety and Depression (HAD) scale consisting of 14 statements (seven for anxiety and seven for depression) (72), and the periodontal evaluation included only CAL and PI. The results showed no association between periodontitis and anxiety when comparing the group with chronic periodontitis and control (p = 0.49; Chi-square test). Based on the data from the study, we calculated the OR (HAD scale: OR = 1.29, CI: 0.63 to 2.706; p = 0.57; Fisher's exact test).

Li et al. (61) evaluated periodontitis using CAL, PPD, and BOP and anxiety using the SAS method. Subjects with periodontitis were divided into three subgroups: moderate, high, severe. Based on the analysis of variance, anxiety indexes of the periodontitis group were higher than those of the control (p < 0.01).

Levin et al. (63) evaluated BOP, PPD, and PI in all participants, and anxiety was measured using the Dental Anxiety Scale (DAS) method. Subjects answered a questionnaire that addressed four dental scenarios for assigning up to five points in each scale's domain (73). From this, it was shown that the levels of dental anxiety were higher in subjects with periodontitis, who were more likely to fear the noise of dental instruments and the application of anesthesia (63). There was a higher percentage of anxiety in subjects with periodontitis (p = 0.036; Chi-square test). Based on the data from the study, we calculated the OR (Self-assessed dental anxiety: OR: 13.16; CI: 5.361–32.5; p < 0.0001; Fisher's exact test).

In the study by Pal et al. (64), CAL and PPD were evaluated, and subjects with periodontitis were categorized into three groups according to the American Academy of Periodontology (AAP): mild (CAL: 1–2 mm), moderate (CAL: 3–4 mm), and severe (CAL ≥ 5 mm). Anxiety was assessed using the BAI method. The findings showed an association between anxiety and the severity of periodontitis (p < 0.001; Chi-square test). Based on the data from the study, we calculated the OR (OR: 7.861; CI: 4.189–14.76; p < 0.0001; Fisher's exact test).

Naghsh et al. (65) study diagnosed periodontitis by considering a CAL ≥ 4 mm or PPD of ≥5 mm. The severity of periodontitis was categorized according to the AAP. BOP and PI were also evaluated in the subjects. Anxiety was assessed using the STAI method. A positive correlation was observed in the Pearson correlation tests between PPD and anxiety score (p < 0.001).

Among all these studies, eight showed a significant association between periodontitis and anxiety (56, 58, 59, 61–65) and three studies did not show significant results for this association (10, 57, 60). However, the different types of methods for assessing anxiety made it impossible to carry out a meta-analysis. The summary of the study characteristics is described in Table 1.

Analysis of methodological quality (Table 2) was carried out for case-control and cross-sectional studies. Some studies have methodological flaws regarding the lack of description of the selection process of the control group (56) and undefined controls (56, 59, 60). Additionally, the absence of description of the rate of non-respondents was observed in all studies (10, 56–65).

Furthermore, Li et al. (61) did not control all confounding factors such as the age and sex of individuals, and Laforgia et al. (62) did not describe the possible confounding factors of the research.

No cross-sectional studies showed any justification for the sample size of the groups evaluated (57, 58, 62, 64, 65).

However, in general, most articles fulfill the criteria of the Newcastle-Ottawa protocol, namely: adequate case definition, the definition of controls, comparability, and evaluation of results. Except for the Vettore et al. (56) study, that presents problems in the control group due to lack of description, and Laforgia et al. (62) that does not describe the justification of the sample size, rate of non-respondents, and control of confounding factors.

The narrative GRADE analysis of evidence profile included only one tool for anxiety evaluation, the State-Trait Anxiety Inventory (STAI). There were other anxiety tests in common between the articles, as the Self-Rating Anxiety Scale (SAS) and the Beck Anxiety Inventory (BAI), but the type of the studies that evaluated them was different (one was a case-control and the other was cross-sectional), which made it impossible to gather the data. Then, the certainty of evidence for the STAI test was divided into case-control (10, 56, 59) and cross-sectional studies (57, 65), and both outcomes presented a very low certainty of evidence. Serious issues with inconsistency were observed due to the lack of homogeneity in statistical evaluation among the studies and imprecision due to the magnitude of effect based on the small number of events (Table 3).