A cross-sectional study assessing the relationship between non-alcoholic fatty liver disease and periodontal disease

The risk factors for non-alcoholic fatty liver disease (NAFLD) progression are not completely known. Porphyromonas gingivalis infection is a risk factor for systemic diseases. We investigated the association of P. gingivalis infection with the risk of non-alcoholic steatohepatitis progression. Here, hematological tests, periodontal examination, and saliva collection were performed for 164 patients with NAFLD. P. gingivalis was identified in saliva using polymerase chain reaction. Hepatic steatosis and stiffness were evaluated using vibration-controlled transient elastography (VCTE) and magnetic resonance imaging. In patients with NAFLD, P. gingivalis positivity (P. gingivalis ratio ≥ 0.01%) in saliva correlated with liver stiffness determined using magnetic resonance elastography (MRE; p < 0.0001). A P. gingivalis ratio of 0.01% corresponds to 100,000 cells/mL and indicates the proportion of P. gingivalis in the total number of bacteria in the oral cavity. Patients with NAFLD and advanced fibrosis on MRE showed significantly elevated endotoxin activity; those who had > 10 periodontal pockets with depths ≥ 4 mm had significantly increased hepatic stiffness on both VCTE and MRE.


Results
According to the sample size calculation, 164 patients who met the eligibility criteria were identified and included between August 2015 and April 2019.
Of the 164 participants, P. gingivalis represented ≥ 0.01% and < 0.01% of all bacteria in the saliva of 45 and 119 patients, respectively.
A P. gingivalis ratio of 0.01% corresponds to approximately 100,000 cells/mL and indicates the ratio of P. gingivalis to the total number of bacteria in the oral cavity quantified by performing polymerase chain reaction (PCR) of the DNA obtained from saliva samples 37,38 . The baseline demographics and disease characteristics of the patient groups with a P. gingivalis ratio ≥ 0.01% and a P. gingivalis ratio < 0.01% in saliva were almost similar ( Table 1). Liver stiffness measured by magnetic resonance elastography (MRE) was significantly higher in the group with a P. gingivalis ratio ≥ 0.01% than in the group with a P. gingivalis ratio < 0.01% in saliva (3.0 ± 1.3 vs. 3.7 ± 1.0, p = 0.0064, Table 1). In addition, there was a significant difference between these groups in terms of the level of the inflammatory marker, ferritin (215 ± 198 vs. 282 ± 142, p = 0.04, Table 1).
Among the 164 study participants, 52 and 112 had less than 10 (3 ± 3 sites) and more than 10 (26 ± 17 sites) periodontal pockets with depths ≥ 4 mm, respectively. The liver stiffness values were significantly higher in the group with > 10 PPD sites than in the corresponding group with < 10 PPD sites (3.6 ± 1.5 vs. 3.0 ± 1.0 kPa, respectively; p < 0.0095; Table 3, Fig. 3a-d). The mean of PPD ≥ 4 mm was 4.0 mm (± 0.8). Moreover, there were significant differences between these groups regarding the periodontal conditions. These data indicate that liver stiffness increases in case of a greater number of deep periodontal pockets.
We also constructed receiver operating characteristic curves for patients using at least 0.01% P. gingivalis in saliva and the number of periodontal pockets with depths ≥ 4 mm as independent variables. When periodontal pockets with depths ≥ 4 mm were observed in at least five sites, the area under the receiver operating characteristic curve, sensitivity, and specificity were 0.73, 0.82, and 0.65, respectively. Similarly, area under the curve of 0.73, sensitivity of 0.41, and specificity of 0.81 were obtained in case the number of periodontal pockets with depths ≥ 4 mm were greater than or equal to 10 ( Fig. 4a,b).
Patients with NAFLD who had > 10 periodontal pockets with depths ≥ 4 mm on periodontal examination had significantly increased liver stiffness on both vibration-controlled transient elastography (VCTE) and MRE. Furthermore, in patients with liver stiffness ≥ 3.4 kPa, there were significant differences in salivary and serum P. gingivalis antibody levels, which may have been related to P. gingivalis' endotoxin.

Discussion
This is a study to demonstrate that the association between NAFLD pathology (especially liver stiffness), P. gingivalis bacteria contents, and oral findings (PPD). This observational study examining the relationship between periodontal disease and NAFLD pathogenesis produced three major findings. First, it demonstrated that the group with ≥ 0.01% P. gingivalis in saliva had a higher MRE-determined liver stiffness than the group with < 0.01% www.nature.com/scientificreports/ P. gingivalis in saliva. P. gingivalis migrates relatively easily from the oral cavity into the blood circulation 39,40 ; therefore, its endotoxin (present in the oral cavity) can reach the liver 41 . Second, it found that the group with MRE-determined liver stiffness values of ≥ 3.4 kPa had higher serum P. gingivalis FDC381 and SU63 antibody levels. P. gingivalis is an obligate anaerobic gram-negative bacillus with strong adhesion; therefore, it forms a biofilm with other gram-negative bacteria associated with periodontal disease. In addition to producing a proteolytic enzyme called gingipains, P. gingivalis expresses lipopolysaccharides (which are endotoxins of the outer bacterial membrane) and is an anaerobic bacterium associated with periodontal disease progression. Inflammation caused by the cytokines and proteins produced during P. gingivalis infection is associated with liver stiffness 42,43 . Furthermore, the histological lesions in NASH are unevenly distributed throughout the liver parenchyma; therefore, liver biopsy errors can result in substantial stratification and staging inaccuracies 44 . We consider it important to perform non-invasive, safe, low-cost, and short-term clinical trials as proof-of-concept studies. Many studies have reported that the steatosis grade could be evaluated using the controlled attenuation parameter [45][46][47][48][49][50] (which is based on the properties of ultrasonographic signals acquired with VCTE) or magnetic resonance imaging (MRI) proton density fat fraction 51,52 (which is an MRI-based method for quantitatively assessing hepatic steatosis and is available for several types of MRI scanners). Moreover, VCTE and MRE have excellent diagnostic abilities for steatosis and fibrosis in patients with NAFLD 53,54 . We consider that non-invasive evaluation of the pathogenesis of NAFLD using VCTE and MRI may replace invasive methods, such as liver biopsy. For this reason, we included non-invasive methods for assessing NASH/NAFLD progression as secondary endpoints and compared the results with those derived from liver biopsies.
Third, the group with ≥ 10 periodontal pockets with depths ≥ 4 mm had increased liver stiffness compared to the group with < 10 periodontal pockets with depths ≥ 4 mm. The presence of more than 10 periodontal pockets with depths > 4 mm suggests high morbidity in patients with periodontal disease 55 . The factor involved in PPD and liver stiffness in this study may have been age, and not sex. With increasing age, the prevalence of periodontal  (34) Hyperuricemia (%) 15 (9) 9 (9) 6 (9)

Liver function
Platelet count 22 (8)  www.nature.com/scientificreports/ disease rises, and periodontal pockets become deeper 56 . However, the average age of the patients was 53.0 years, and age was not associated with the other examined factors in this study. Inflammation caused by the cytokines and proteins produced in response to P. gingivalis infection is associated with liver stiffness 42,43 . Therefore, the amount of bacteria in the oral cavity and serum antibody titers were examined in this cross-sectional study, and the relationship between periodontal disease and liver stiffness in patients with NAFLD was evaluated.
NAFLD is common in patients with metabolic abnormalities, such as type 2 diabetes and obesity; its prevalence is rapidly increasing globally with the growing obese population 5 . The main risk associated with this disease is that it progresses from liver cirrhosis to liver cancer over a long time without manifestation of accompanying subjective symptoms, and mortality due to cerebral or myocardial infarction is also high. It is thought that the pathology in patients with NASH progresses as periodontopathic bacteria enter the blood, leading to  Table 3. Baseline characteristics of the study population. ALP alkaline phosphatase, ALT alanine aminotransferase, AST aspartate aminotransferase, BMI body mass index, BOP bleeding on probing, CAP controlled attenuation parameter, CAL clinical attachment level, DPP4 dipeptidyl peptidase-4, EAA endotoxin activity assay, GGT gamma-glutamyl transferase, HDL-C high-density lipoprotein cholesterol, HOMA-R homeostasis model assessment-estimated insulin resistance, hsCRP high-sensitivity C-reactive protein, IMT (R: right, L: left) intima-media thickness (R, L), LDL-C low-density lipoprotein cholesterol, LSM liver stiffness measurement, MRE magnetic resonance elastography, MRI magnetic resonance imaging, PDFF proton density fat fraction, PPD periodontal pocket depth, T.Bil total bilirubin, Tcho total cholesterol, TG triglycerides, VCTE vibration-controlled transient elastography. a Data are reported as means (SDs) or numbers (percentages). The p-values were determined using the t-test. www.nature.com/scientificreports/ inflammation in the fatty liver. Therefore, it is necessary to actively interfere with the link between periodontal disease and NAFLD. The strength of this study lies in the use of non-invasive MRI (in place of liver biopsy) to assess changes in liver fat content. Importantly, this non-invasive method revealed a seemingly mutual relationship between P. gingivalis and liver stiffness.
The study has a certain limitation. It would be important to differentiate NASH from NAFLD in the patients enrolled in this study; however, currently, biopsy is not used to diagnose and differentiate liver tissue.
In summary, patients with NAFLD who had P. gingivalis infection and deeper periodontal pockets had higher liver stiffness values than those who did not have periodontal disease.
Furthermore, the serum antibody titers for P. gingivalis FDC381 and SU63 were high in the patient group with increased liver stiffness, and logistic regression analysis suggested that periodontal disease is associated with liver stiffness. These results indicate that periodontal disease may be associated with NAFLD. As this was a cross-sectional study, a causal relationship could not be ascertained. Additional randomized controlled trials are needed to investigate the relationship between periodontal disease and NAFLD, and further investigation of patients with NAFLD who do not have periodontal disease is needed. As this study showed that liver fibrosis and periodontal disease are likely to be related, future interventional studies with treatment for fibrosis as an endpoint will be necessary. Table 4. Factors associated with NAFLD in advanced fibrosis. AST aspartate aminotransferase, BOP bleeding on probing, IMT intima-media thickness, NAFLD non-alcoholic fatty liver disease, PPD periodontal pocket depth.

Covariate
Odds ratio 95% confidence interval p-value   Each participant was examined to assess the amount of periodontal disease-causing bacteria (e.g., P. gingivalis, T. fosythensis, and T. denticola) present in oral cavity, the degree of infection, and periodontal disease severity. All examinations were performed by dentists affiliated with Kanagawa Dental University Hospital and Kanagawa Dental University Yokohama Clinic. The amount of P. gingivalis in saliva was measured using quantitative PCR, and the infection level was evaluated using serum immunoglobin G (IgG) antibody titer tests for P. gingivalis FDC381 and SU63 based on enzyme-linked immunosorbent assays. The severity of periodontal disease was determined by assessing the PPD, clinical attachment level, level of gingival BOP at six sites per tooth using a calibrated periodontal probe (HU-FRIEDY PQ-OW Thin Williams, Hu-Friedy , Chicago, IL, USA), and the stability of the teeth. The patients with NAFLD recruited in this study met the inclusion and exclusion criteria listed in Supplementary Tables S1 and S2, respectively. Fatty liver, steatosis grade, and fibrosis stage were evaluated using non-invasive methods such as ultrasonography (VCTE) and MRI.

Registration.
For patients judged to be eligible, the principal investigator or collaborator filled out a patient registration form with the necessary information and sent these data to the Patient Enrollment Center (Yokohama City University). The Patient Enrollment Center confirmed the patient's eligibility based on the registration form, registered the patient, and informed the main researcher or collaborator of the patient's identification and assignment numbers. All participants provided informed consent to participate in the study prior to the commencement of the study.

Sample size calculation.
A retrospective analysis of patients with NAFLD and periodontitis (P. gingivalis ratio < 0.01% vs. P. gingivalis ratio ≥ 0.01%) in Yokohama City University Hospital showed a mean liver stiffness of 2.8 and 3.6 in the P. gingivalis < 0.01% and P. gingivalis ≥ 0.01% groups, respectively, on MRE. We decided to calculate the number of patients required for a proper analysis of the variance t-test based on these data. Considering the liver stiffness (determined with MRE) in the P. gingivalis < 0.01% and P. gingivalis ≥ 0.01% groups to be 2.8 and 3.6, respectively, with a common SD of 1.8, a total of 159 patients were needed to reach an 80% statistical power with a two-sided significance level of 5%. To compensate for dropouts, we proposed an increase in the number of patients to 164.
Statistical analyses. In this study, statistical analyses were performed mainly for the following items. As primary analysis, stratification analysis was performed using the P. gingivalis bacterial content, MRE, and periodontal pockets as indexes. At the secondary endpoint, the P. gingivalis IgG antibody titer in the blood was required for each participant, and the paired Wilcoxon signed-rank test was performed. Data are presented as means ± SDs or numbers (percentages). The significance level was set at p < 0.05. We used statistical software for all analyses (SAS version 9.4, SAS Institute, Cary, NC, USA).
Trial steering and data monitoring committees. The pilot committee consisted of three people appointed by independent clinical and basic researchers (general and palliative care specialists and statisticians at Yokohama City University School of Medicine). They provided overall supervision and ensured all registered examinations. These investigators were anonymous and randomly selected. The Independent Data Monitoring Committee comprised two members of staff from the Department of Biostatistics, Yokohama City University School of Medicine. The management team monitored the monthly progress of patients and data with phone, email, and/or web conferencing. If the oversight committee determined that field monitoring was necessary, a member of the Independent Data Monitoring Committee visited the site for face-to-face monitoring.

Ethics declarations. This study involves human participants and was approved by the Kanagawa Dental
University Committee for Research Screening (Approval No: 323). All procedures were performed in accordance with the Declaration of Helsinki.
Consent to participate/Consent to publish. All participants provided informed consent to participate in the study prior to initiation of the study. Participants also signed a form indicating their agreement to the presentation of their data at conferences and in papers in a manner that does not disclose their personal information.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information files.