Association of circulating leptin and adiponectin with periodontitis: a systematic review and meta-analysis

The inclusion criteria for the studies comparing the difference in serum levels of leptin and adiponectin between patients with periodontitis and periodontally healthy individuals were as follows: (1) the studies were case-controlled, cross-sectional, prospective, or clinical trials; (2) the periodontitis groups consisted of patients diagnosed with periodontal disease, including chronic periodontitis (CP) and aggressive periodontitis (AP). Present study employed the definition of periodontitis as: at least 1 sites with probing pocket depth (PPD) ≥4 mm [34]; (3) the healthy groups consisted of periodontally healthy individuals; and (4) serum levels of leptin and adiponectin were detected and provided in both the periodontitis and healthy groups.

The inclusion criteria for the studies evaluating the change in circulating leptin and adiponectin after periodontal treatment were as follows: (1) the studies were randomized controlled trials, nonrandomized controlled trials, or prospective; (2) patients with periodontitis were enrolled; (3) periodontal treatments such as scaling and root planning and periodontal surgeries with or without antibiotics administration were applied; and (4) serum levels of leptin and adiponectin of the patients with periodontitis at baseline and after the periodontal treatments were provided.

The exclusion criteria were as follows: (1) the studies were review articles, case reports, letters, or conference abstracts; (2) studies with insufficient data for the statistical analyses; and (3) those with repeated data.

Pubmed, Embase, Web of Science, and Cochrane Library up to September 2016 were systemically searched to identify the pertinent studies. The search strategies are presented in Table 1. In addition, the reference lists of the selected manuscripts and related reviews were also manually searched and screened for a comprehensive search result.

According to the eligibility criteria, titles and abstracts were screened first and the full-text paper screen was conducted next. The results were screened by two authors (Xin Chen and Ling Zhang) independently, and a third author (Yuting He) was consulted if any discrepancy existed.

The following data were extracted from the included studies by two authors (Junfei Zhu and Ling Zhang) independently, and discrepancies were resolved through discussion: (1) Name of the first author and year of publication; (2) study design; (3) country and ethnicity; (4) group size; (5) body mass index (BMI); (6) age; (7) outcomes; (8) systemic conditions; (9) smoking status; (10) posttherapy time; and (11) serum levels of leptin and/or adiponectin.

The methodological quality of the included studies was evaluated by two authors (JF Zhu and L Zhang) independently, and discrepancies were resolved through discussion. The quality of the included cross-sectional studies was assessed using the quality assessment form for cross-sectional studies recommended by the Agency for Healthcare Research and Quality (AHRQ). A total of 11 items were involved in this form, and 1 point could be achieved if the item was reflected in the study. No points could be achieved if the item was not considered or unclear. The final scores ranged from 0 to 11. Studies with scores 0–4, 5–8, and 9–11 were considered as of low, moderate, and high quality, respectively [35, 36]. In addition, for the studies evaluated the changes in serum levels of leptin and adiponectin after periodontal treatment, the included double-arm clinical trials and randomized controlled clinical trials were treated as single-arm clinical trial designs. Therefore, the Methodological Index for Nonrandomized Studies (MINORS) was applied for assessing the quality of the included clinical trials [37]. A total of 12 items were involved in MINORS, and 0–2 points could be achieved in each of the items. The first eight items were designed for the studies without a control group,and the other four items are the additional criteria for comparative studies. As in present quality assessment we only evaluated the quality of single-arm designs, the first 8 items were employed. The final scores ranged from 0 to 16. Studies with scores 0–5, 6–10, and 11–16 were considered as of low, moderate, and high quality, respectively [38].

Four models were considered in the present meta-analysis: serum levels of leptin in patients with periodontitis versus healthy individuals (L1), serum levels of adiponectin in patients with periodontitis versus healthy individuals (A1), serum levels of leptin in patients with periodontitis before versus after periodontal treatment (L2), and serum levels of adiponectin in patients with periodontitis before versus after periodontal treatment (A2). The data on serum levels of leptin and adiponectin were presented as mean (M) ± standard deviation (SD). If the outcomes were provided only as median (minimum – maximum) [39‐41], the results were shifted to the form of M ± SD according to the estimation method reported by Hozo et al. [42]. The standard mean difference (SMD) and corresponding 95% confidence interval (CI) were applied to estimate the association between serum levels of leptin and adiponectin and periodontitis. Heterogeneity was estimated by chi-square and I; a P value <0.05 and I ² > 50% were considered significant heterogeneity [43], and then the random-effects model was used; otherwise, a fixed-effects model was used to assure statistical efficiency. Subgroup analyses were conducted according to the type of periodontitis, BMI, smoking status, systemic conditions, and types of periodontal treatment. The following characteristics were included as covariates in the meta-regression to explore the potential sources of heterogeneity: type of periodontitis, BMI, smoking status, and presence of systemic disease. A characteristic was considered as the source of heterogeneity if P <0.05. Moreover, Galbraith plots were conducted to investigate the heterogeneity. Sensitivity analyses were performed to test the robustness of the results. Publication bias was measured by Begg’s and Egger’s linear regression tests. A significant publication bias was considered if the P value was <0.05. All statistical analyses were processed using the STATA 12.0 software.

A total of 399 manuscripts were yielded using the search strategy. After deleting the duplicates, full-text screens were conducted and 25 studies were identified to be eligible. The process of study selection and the reasons for exclusion are listed in Fig. 1.

Of the 25 included studies, all 16 studies focusing on the difference in serum levels of leptin and/or adiponectin between patients with periodontitis and healthy individuals employed the cross-sectional design [16, 30‐32, 40, 41, 44‐53]. Besides this, nine clinical trials detecting the serum levels of leptin and adiponectin in patients with periodontitis before and after periodontal treatments were included [28, 29, 39, 54‐59]. The clinical trials performed by Duzagac et al., Purwar et al., and Shimada et al. also enrolled the periodontally healthy groups and provided the pretherapy data on the serum levels of leptin and adiponectin in patients with periodontitis and healthy individuals [54, 57, 58]. The studies performed by Altayet al., Duzagacet al., Goncalveset al, Mendoza-Azpuret al, and Zimmermann et al. provided the data on subjects with BMI <30 and BMI ≥30 separately [31, 32, 39, 54‐56], and the study performed by Zeigler et al. provided only the data on obesity subjects. The study performed by Ay et al. enrolled both the CP and aggressive periodontitis (AP) groups [40]. The study performed by Gundala et al. enrolled both the systemically healthy group and the acute myocardial infarction (AMI) group [47]. The study performed by Davies et al. provided the data on male and female subjects separately [44]. For the included cross-sectional studies, the AHRQ scores ranged from 5 to 7. The study performed by Gangadhar et al. was the only cross-sectional study considered to be of low quality (4 points) [46]. Other included cross-sectional studies were all considered to be of moderate quality. All of the included clinical trials were considered to be of high quality. The characteristics and quality assessment of the included studies are presented in Table 2.

In the L1 model, higher serum levels of leptin were observed in the overall and subgroup analyses of BMI <30. The subgroup analysis of BMI ≥30 did not show significant results. The overall and subgroup analyses of the L1 model demonstrated significant heterogeneity. In the A1 model, the overall and subgroup analyses showed a significantly decreased serum level of adiponectin in patients with periodontitis, except for the subgroup of BMI ≥30. In the L2 model, the overall and subgroup analyses showed no significant change in serum levels of leptin in patients with periodontitis after periodontal treatment. In the A2 model, the overall and subgroup analyses of BMI <30 showed a significantly elevated level of serum adiponectin in patients with periodontitis after periodontal treatment. However, no significant result was obtained by the subgroup analysis of BMI ≥30 and the subgroup analyses based on the methods of periodontal treatment and systemic conditions. The results of overall and subgroup analyses are summarized in Table 3, and the forest plots of the subgroup analyses of BMI <30 are presented in Fig. 2.

Because of the presence of heterogeneities in the subgroup of BMI <30 under all of the four models and the inconsistency of the results between the subgroups of BMI <30 and BMI ≥30 under L1, A1, and A2 models, the meta-regression analyses and Galbraith plots were conducted to investigate the heterogeneity in the subgroup analyses of BMI <30. In addition, the sensitivity analyses were performed to test the stability of the results.

Considering the limitation in the number of studies included in L2 and A2 models (n < 10), the meta-regression analyses were performed only in the L1 and A1 models. The results of meta-regression based on the covariates including the type of periodontitis, BMI, systemic conditions, and smoking status failed to find the source of heterogeneity in the subgroup analyses of BMI <30 (P > 0.05).

Further, the Galbraith plots showed that the studies performed by Shi et al., Purwar et al., Mendoza-Azpur et al., Thanakun et al., and Karthikeyan et al. (L1) [30, 31, 49, 51, 52]; Mendoza-Azpur et al. and Sete et al. (A1) [16, 31]; and Purwar et al. (L2) [57] were located outside of the two lines in the Galbraith plots of the subgroups of BMI <30. After removing the outliers, the heterogeneities effectively decreased, and the results were not materially changed in the subgroups of BMI <30 (L1: SMD = 0.796, 95% CI: 0.572–1.020, P = 0.000; I ² = 40.9, P = 0.095; A1: SMD = −0.251, 95% CI: −0.408 to −0.095, P = 0.002; I ² = 13.6, P = 0.323; L2: SMD = −0.162, 95% CI: −0.417 to 0.094, P = 0.215; I ² = 0.0, P = 0.556). Additionally, in the L1 model, removing the outlier also eliminated the overall heterogeneity, and the overall result was not materially changed (SMD = −0.175, 95% CI:−0.383 to 0.033, P = 0.100; I ² = 0.0, P = 0.671).

The sensitivity analyses were conducted by removing a single study each time and evaluating the influence of each study on the result. The sensitivity analyses proved the robustness of the results of the BMI <30 subgroups under the L1, A1, and L2 models. However, in the A2 model, the sensitivity analysis showed that the study performed by Sun et al. qualitatively changed the pooled SMD [59]. After this study was removed, the results were changed to the insignificant level, and the heterogeneity was eliminated in both overall and BMI <30 analyses (overall: SMD = 0.109, 95% CI: 0.113–0.330, P = 0.336; I ² = 0.0, P = 0.839; BMI <30: SMD = 0.065, 95% CI:−0.238 to 0.369, P = 0.673; I ² = 0.0, P = 0.612). The Galbraith plots and sensitivity analyses are presented in Figs. 3 and 4.

The Begg’s and Egger’s linear regression tests demonstrated the presence of a publication bias in the A2 model (P = 0.042) (Table 2).