Abstract
Objective
There are conflicting reports on the relationship between vitamin D and periodontal disease. Our research is intended to further analyse the association between serum 25(OH)D3, a vitamin D precursor and periodontal disease based on a large national survey sample in Japan.
Methods
We downloaded the 2009–2018 National Health and Nutrition Examination Survey (NHANES) cycle, which included a total of 23,324 samples. Logistic regression of factors influencing perioral disease including periodntal disease, and subgroup logistic regression were performed to analyse the relationship between serum vitamin D and perioral disease, using WTMEC2YR as weights for regression analysis. Then machine learning model–based prediction of perioral disease onset was performed, and the machine learning algorithms used included boost tree, artificial neural network, AdaBoost, and random forest.
Results
We evaluated the vitamin D, age, sex, race, education, marriage, body mass index, ratio of family income to poverty (PIR), smoking, alcohol consumption, diabetes, and hypertension as variables in the included samples. Vitamin D was negatively associated with perioral disease; compared with Q1, the odds ratios and 95% CI were 0.8 (0.67–0.96) for Q2, 0.84 (0.71–1.00) for Q3, and 0.74 (0.6–0.92) for Q4 (P for trend <.05), respectively. The results of the subgroup analysis showed that the effect of 25(OH)D3 on periodontal disease was more pronounced in women younger than 60 years. Based on the accuracy and receiver operating characteristic curve, we concluded that a boost tree was a relatively good model to predict periodontal disease.
Conclusions
Vitamin D might be a protective factor for periodontal disease, and boost tree analysis we emplyed was a relatively good model to predict perioral disease.
Key words: Serum vitamin D, Periodontal disease, Machine learning, NHANES
Introduction
Periodontal infection is a common disease that mainly affects young people. It is characterised by rapid destruction of periodontal tissue and can lead to early tooth loss.1,2 The aggressiveness and onset of the disease depend on many factors, including the patient's susceptibility to infection. Localised periodontal infection is known to cause a systemic response in the host and leads to elevated levels of systemic inflammatory markers, including white blood cell count, C-reactive protein, and interleukin (IL)-6 levels, some of which are predictive markers of systemic disease.3
Vitamin D deficiency conditions have been shown to be associated with a variety of diseases, including oral health disorders.4,5,6 Inadequate sunlight exposure may accelerate the onset of some of these disorders, which may be due to impaired vitamin D synthesis.7,8 The beneficial effects of vitamin D on oral health are not limited to its direct effects on tooth mineralisation but are also through its anti-inflammatory function and its ability to stimulate the production of antimicrobial peptides.9,10
However, previous studies have reported little association between serum 25(OH)D3 levels and periodontal status, and the relationship with periodontal disease was analysed by measuring 25(OH)D3 levels in current nonsmoking and nondiabetic adults aged 30 to 49 years.11 It was shown that low serum 25(OH)D3 levels do not appear to be a risk factor for infectious periodontal disease.
There is still conflicting evidence about the relationship between vitamin D and periodontal disease. Therefore, we performed further analysis of the relationship between serum 25(OH)D3 and periodontal disease based on data from a large sample.
Methods
Study design
We obtained data from the National Health and Nutrition Examination Survey (NHANES).12,13 More details about the NHANES programme and study procedures are available online (https://wwwn.cdc.gov/nchs/nhanes/).14,15,16,17 Demographic characteristics and lifestyle, including age, sex, race/ethnicity, smoking and alcohol use status, measurement of anthropometric data (including height, weight, and waist circumference) and blood pressure as part of a physical examination, and blood samples for vitamin D testing were obtained in this survey.
This study used the 2009–2018 NHANES cycle and included a total of 23,324 samples containing data on demographics, lifestyle habits, disease history, household economics, and perioral disease that we needed to study.
Serum vitamin D and covariates
The US Centers for Disease Control and Prevention use high-performance liquid chromatography tandem mass spectrometry (HPLC-MS/MS) to quantify 25-hydroxyvitamin D3 (25OHD3), 3-epi-25-hydroxyvitamin D3 (epi-25OHD3), and 25-hydroxyvitamin D2 (25OHD2) in human serum. The analytes were typically separated chromatographically on 1 of 3 pentafluorophenyl columns. The mobile phase composition of the optimised chromatogram was slightly different for the 3 columns, but the methanol content in water ranged between 69% and 72%. The composition of the solution added to the serum prior to extraction, the solution used for reconstitution, and the needle wash should match the composition used for the mobile phase.
Co-variates term, household and demographic questionnaires were administered at home by trained interviewers using the Computer Assisted Personal Interviewing (CAPI) system. Minors aged 16 and older and emancipated minors were interviewed directly. The CAPI system was programmed with built-in consistency checks to reduce data entry errors. CAPI also used an online help screen to help interviewers define key terms used in the questionnaire. After collection, NHANES field office staff reviewed the interview data to ensure the accuracy and completeness of the items selected. Interviewers were required to record interviews on a regular basis, and recorded interviews were reviewed by National Center for Health Statistics staff and interviewer supervisors.
Periodontal disease
The oral health questionnaire section provides personal interview data on oral health topics and contains reports of periodontal disease. This section was administered at home prior to the medical examination using the CAPI (Interviewer Administration) system. The data about periodontal disease were self-reported, which was useful for population-wide studies of periodontitis with limitations inherent in basing assessment of periodontitis.18
Statistical analysis
All the analyses used the appropriate data weight of NHANES examination samples for 10 years, considering the complex research design and nationally representative estimate. In this study, WTMEC2YR was used as the weight of the regression analysis, and the logistic regression and subgroup logistic regression of the factors affecting perioral disease were carried out. After analysing the relationship between Xueqing vitamin D and perioral disease, we predict the incidence of perioral disease based on the machine learning model. The machine learning algorithms used include boost tree, artificial neural network, AdaBoost, and random forest. Based on recommendations about machine learning applied to the diagnosis of diseases, we trained all 4 machine learning models in the same way to make comparisons amongst the models. The original dataset was cut into 7:3, in which 70% data was used for model training and the other 30% data was for validation of the model. Receiver operator characteristic (ROC) curves were conducted for the validation of clinical diagnosis machine learning models.19
All analyses were performed using R4.2.2. P < .05 was considered statistically significant.15,20
Results
The basic characteristics can be found in Table 1. We counted the included samples, grouped by the presence or absence of perioral disease, for vitamin D, age, sex, race, education, marriage, body mass index (BMI), ratio of family income to poverty (PIR), smoking, alcohol consumption, diabetes mellitus, and hypertension. There were some differences in vitamin D levels between the 2 groups, 64.5 ± 27.9 in the group with perioral disease and 68.2 ± 29.7 in the group without perioral disease, but the differences in other covariates were not significant.
Table 1.
Characteristics of the included population.
| Periodontal disease (n = 4299) | No periodontal disease (n = 19,025) | |
|---|---|---|
| Vitamin D, nmol/L | 64.5 (27.9) | 68.2 (29.7) |
| Age, y | 53.3 (13.3) | 54.9 (15.4) |
| Sex | ||
| Male | 2197 (51.1%) | 9076 (47.7%) |
| Female | 2102 (48.9%) | 9949 (52.3%) |
| BMI, kg/m2 | ||
| Normal (<25) | 1006 (24.5%) | 4714 (26.2%) |
| Overweight (25≤BMI<30) | 1841 (44.8%) | 7202 (40.1%) |
| Obesity (≥30) | 1263 (30.7%) | 6063 (33.7%) |
| Race | ||
| Mexican American | 670 (15.6%) | 2568 (13.5%) |
| Other Hispanic | 473 (11.0%) | 1970 (10.4%) |
| Non-Hispanic White | 1658 (38.6%) | 7730 (40.6%) |
| Non-Hispanic Black | 893 (20.8%) | 4105 (21.6%) |
| Other | 605 (14.1%) | 2652 (13.9%) |
| Education | ||
| Less than 9th grade | 493 (11.5%) | 2119 (11.1%) |
| 9th-11th grade | 655 (15.2%) | 2476 (13.0%) |
| High school | 1007 (23.4%) | 4135 (21.7%) |
| Some college | 1252 (29.1%) | 5342 (28.1%) |
| College graduate or above | 886 (20.6%) | 4922 (25.9%) |
| Marital status | ||
| Married | 2237 (52.0%) | 10841 (57.0%) |
| Widowed | 322 (7.49%) | 1901 (9.99%) |
| Divorced | 644 (15.0%) | 2336 (12.3%) |
| Separated | 211 (4.91%) | 659 (3.46%) |
| Never married | 537 (12.5%) | 2094 (11.0%) |
| Living with partner | 346 (8.05%) | 1177 (6.19%) |
| Refused | 2 (0.05%) | 15 (0.08%) |
| PIR | 2.30 (1.59) | 2.62 (1.64) |
| Smoking | ||
| Yes | 2321 (54.0%) | 8219 (43.2%) |
| No | 1975 (45.9%) | 10794 (56.7%) |
| Alcohol | ||
| Yes | 793 (23.6%) | 2324 (16.6%) |
| No | 2572 (76.4%) | 11659 (83.3%) |
| Diabetes | ||
| Yes | 843 (19.6%) | 2836 (14.9%) |
| No | 3301 (76.8%) | 15658 (82.3%) |
| Borderline | 148 (3.44%) | 523 (2.75%) |
| Hypertension | ||
| Yes | 150 (3.92%) | 568 (3.41%) |
| No | 3680 (96.1%) | 16088 (96.6%) |
PIR, ratio of family income to poverty.
The relationship between serum vitamin D and perioral disease was analysed by constructing 3 models (Table 2). Model 1 included only serum vitamin D as independent variables, and model 2 included serum vitamin D, age, sex, education, ethnicity, PIR, smoking, and alcohol consumption as independent variables. Model 3 added BMI, history of hypertension, and history of diabetes mellitus as independent variables to model 2. Setting the first quartile as the reference, the relationship amongst the 4 quartiles of serum vitamin D and perioral disease was then analysed, and it was seen that all odds ratios (ORs) were negative and significant, indicating that vitamin D has a protective effect on perioral disease. The P value for trend was less than .05, indicating that there was a dose–response relationship between vitamin D and perioral disease under all 3 models. It can be seen that the relationship between the 2, with stable results, is statistically significant.
Table 2.
Association between serum 25(OH)D3 and periodontal disease
| Serum 25(OH)D3 | Model 1 | Model 2 | Model 3 |
|---|---|---|---|
| Q1 | Reference | Reference | Reference |
| Q2, OR (95% CI) | 0.78 (0.68–0.89)* | 0.81 (0.69–0.94)* | 0.8 (0.67–0.96)* |
| Q3, OR (95% CI) | 0.76 (0.68–0.86)* | 0.83 (0.71–0.96)* | 0.84 (0.71–0.99)* |
| Q4, OR (95% CI) | 0.63 (0.55–0.72)* | 0.73 (0.62–0.87)* | 0.74 (0.6–0.92)* |
| P for trend | <.05 | <.05 | <.05 |
P < .05, model 1 = no adjust. Model 2 = as model 1 plus adjusted for sex, age (years, continuous), education (less than high school, high school graduate, some college and above), race (non-Hispanic White, non-Hispanic Black, Mexican American, other), self-reported alcohol status (yes and no) and self-reported smoking status (yes and no), and PIR. Model 3 = model 2 plus adjusted for body mass index, self-reported hypertension (yes and no), and self-reported diabetes (yes and no). The OR (95% CI) of reference is 1.
PIR, ratio of family income to poverty.
To further analyse the relationship between serum vitamin D and perioral disease, we sliced the dataset by sex and age and then performed further subgroup logistic regression (Table 3). In the sex subgroup, it can be seen that the absolute value of OR is greater in women than in men, and the absolute value of OR is greater in the age <60 years group than in the ≥60 years group. Women younger than 60 years old are more likely to be affected by serum vitamin D levels whether they have perioral disease or not.
Table 3.
Subgroup analysis of association between serum 25(OH)D3 and periodontal disease
| Serum 25(OH)D3 | Male | Female | Age <60 y | Age ≥60 y |
|---|---|---|---|---|
| Q1 | Reference | Reference | Reference | Reference |
| Q2, OR (95% CI) | 0.78 (0.6–0.91)* | 0.8 (0.63–0.93)* | 0.75 (0.6–0.93)* | 0.81 (0.58–0.93)* |
| Q3, OR (95% CI) | 0.92 (0.71–0.98)* | 0.82 (0.64–0.94)* | 0.8 (0.66–0.97)* | 0.95 (0.66–0.95)* |
| Q4, OR (95% CI) | 0.84 (0.64–0.91)* | 0.63 (0.46–0.87)* | 0.72 (0.56–0.92)* | 0.73 (0.5–0.96)* |
| P trend | <.05 | <.05 | <.05 | <.05 |
P < .05. The OR (95% CI) of reference is 1.
In order to help patients to better prevent perioral disease in clinical applications, we used the demographic data, lifestyle habits, disease history, and serum vitamin D levels of the sample to predict the occurrence of perioral disease. We used 4 machine learning algorithms including boost tree, artificial neural network, AdaBoost, and random forest to evaluate the accuracy of the models (Figure 1). We calculated the accuracy and kappa value of each model after prediction, and it can be seen that boost tree has the highest accuracy and kappa value.
Fig. 1.
Evaluation of 4 machine learning models for periodontal disease prediction. BTmodel, boost tree; ANNmodel, artificial neural network; ADmodel, AdaBoost; RFmodel, random forest.
Because dataset of perioral disease is binary, we analysed the receiver operating characteristic (ROC) curve of the predicted values and calculated the area under the curve (AUC) to further compare the strengths and weaknesses of each model (Figure 2). The boost tree model has a maximum AUC value of 0.63 ± 0.02. Although the boost tree model had a poor prediction effect, it had a better prediction than the other 3 machine learning models. In the future, we would try to explore some other more accurate machine learning models.
Fig. 2.
ROC and AUC of 4 machine learning models for periodontal disease prediction. A, Boost tree. B, Artificial neural network. C, AdaBoost. D, Random forest. ROC, receiver operating characteristic; AUC, area under curve.
To further analyse the contribution of input variables for predicting perioral disease, we conducted variable importance plots (Supplementary Figure 1). The plot showed that vitamin D contributed the most in boost tree, PIR contributed most in both artificial neural network and AdaBoost, and age contributed most in random forest. In addition, we conducted ROC curves without vitamin D and the results showed that the AUC of ROC changed from 0.63 to 0.6 in boost tree, 0.61 to 0.57 in artificial neural network, 0.59 to 0.59 in AdaBoost, and 0.59 to 0.57 in random forest. These changes indicated that the input of vitamin D is meaningful to the performance of machine learning models (Supplementary Figure 2). To further analyse the difference of ROC curves amongst 4 machine learning models, we listed the details of ROC comparisons in Supplementary Table 1. In sensitivity, negative prediction, false positive rate, and F1 value, the AdaBoost model had the relatively best performance amongst 4 machine learning models.
Discussion
The results of this study, based on a large sample of 23,324 cases, analysed the relationship between serum 25(OH)D3 and periodontal disease and confirmed that serum 25(OH)D3 had a likely protective effect on periodontal disease. In addition, a dose–response relationship was noted, with increasing serum 25(OH)D3 levels resulting in greater protection against periodontal disease. At the same time, in all the models, Q3 had the lowest protectiveness amongst Q2 to Q4 serum 25(OH)D3 level, which indicated a nonlinear relationship between vitamin D and periodontal disease. Li et al21 indicated a threshold effect and potential nonlinear link between periodontitis and 25(OH)D metabolites, and this is consistent with our results. Regarding the underlying mechanism of complex and nonlinear association between vitamin D and periodontal disease, 25(OH)D3 might modulate periodontal inflammation and bone absorption by inhibiting the production of IL-8 and Monocyte chemoattractant protein-1.22 The results of the subgroup analysis showed that the effect of 25(OH)D3 on periodontal disease was more pronounced in women younger than 60 years. Previous reviews have shown inconsistent results from studies on different vitamin D levels in men and women. In general, women are at higher risk of vitamin D deficiency in some geographic areas. Compared with men, women are more likely to experience functional impairments and abnormalities of the skeletal system associated with low vitamin D levels.23 The relationship between serum 25(OH)D3 and periodontal disease need to be further investigated. A few postulated mechanisms of this realtionship that have been reported thus far are provided below.
Some studies found that CYP27B1 expression in gingival tissue was higher in the periodontitis group than in the control group, providing new evidence for the involvement of the vitamin D pathway in periodontal immune defense.22,24 The vitamin D pathway has been shown to be present in human gingival fibroblasts and human gingival epithelium. This pathway might be involved in periodontal immune defense; CYP27B1 expression had been detected in human gingival fibroblasts in vivo, and this expression may be induced by periodontitis.25,26 The biological function of vitamin D is mediated by its strong antimicrobial, anti-inflammatory, and host-regulatory properties. Experimental periodontitis models involving targeted deletion of 1α-DD hydroxylase, the enzyme responsible for converting inactive substrates to active, 25(OH)2D 3, show increased alveolar bone loss and gingival inflammation.24 Vitamin D receptor (VDR) gene polymorphisms have also been associated with increased severity of periodontitis. In addition to its role in bone and calcium homeostasis, the ability of vitamin D to modulate the adaptive immune response by selectively stimulating specific T-beta helper cell subsets could promote an environment conducive to inflammatory regression. Monocytes/macrophages can be induced to secrete molecules with strong antibiotic effects.25
The biomarker 25(OH)D3 reflected the contribution of all 3 sources of vitamin D—diet, supplements, and sunlight—and 25(OH)2D 3 induced the expression of the antimicrobial peptide LL-37 and innate immune mediators in cultured human gingival epithelial cells. Also, 25(OH)2D 3 activation of the VDR induced the expression of cAMP and defensin, 2 peptides with antimicrobial activity.27, 28, 29, 30
It was recently reported that low values of circulating vitamin D were associated with increased levels of chronic periodontitis, and although there was no causative relationship between levels of vitamin D and chronic periodontitis.21,31
In addition, 4 machine learning algorithms were trained for predicting periodontal disease including boos`t tree, artificial neural network, AdaBoost, and random forest. Boost tree had the highest accuracy and kappa value. Machine learning is essentially a data-driven approach that can simultaneously navigate information far beyond human capabilities and therefore may make use of larger amounts of data rather than smaller ones.32 To better explain the machine learning algorithms, we also conducted variable importance analysis and the results showed that vitamin D played an important role in the boost tree model. In addition, the work of machine learning should be combined with clinical data and knowledge. To further make the variable contribution clear, we conducted variable importance plots and the results showed that vitamin D contributed the most in the Boost tree model.
In conclusion, based on a large sample of 23,324 cases, we found the possible protective effects of serum 25(OH)D3 on periodontal disease, especially in women younger than 60 years, and we trained a boost tree model with relatively good accuracy and effectiveness to predict the occurrence of periodontaldisease.
One strength of this study is that we constructed the machine learning based on the results obtained from logistic regression analyses. The other strengths include the large scale sample size, continuous 10-year survey, and adequacy of covariates.
Our study has some limitations. First, the relationship between serum 25(OH)D3 and periodontal disease could not be firmly ascertained as the samples included were from a cross-sectional survey, not a cohort study. Second, the included covariates were based on retrospective survey data and the effect of other possible confounding covariates could not be evaluated. Third, the sample included in this study that spanned a historical period of 2009-2018 and the current ground situation could be different. Finally, the machine learning algorithms we used had limited prediction performance. Better predictive models should be used in the future.
Conflict of interest
None disclosed.
Footnotes
Supplementary material associated with this article can be found in the online version at doi:10.1016/j.identj.2023.06.004.
Appendix. Supplementary materials
REFERENCES
- 1.Uwitonze AM, Murererehe J, Ineza MC, et al. Effects of vitamin D status on oral health. J Steroid Biochem Molec Biol. 2018;175:190–194. doi: 10.1016/j.jsbmb.2017.01.020. [DOI] [PubMed] [Google Scholar]
- 2.Zhang X, Meng H, Sun X, et al. Elevation of vitamin D-binding protein levels in the plasma of patients with generalized aggressive periodontitis. J Periodont Res. 2013;48(1):74–79. doi: 10.1111/j.1600-0765.2012.01505.x. [DOI] [PubMed] [Google Scholar]
- 3.Liu K, Han B, Hou J, Zhang J, Su J, Meng H. Expression of vitamin D 1α-hydroxylase in human gingival fibroblasts in vivo. PeerJ. 2021;9:e10279. doi: 10.7717/peerj.10279. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Soraya G, Liu J, Javed A, et al. The vitamin D response element in the distal osteocalcin promoter contributes to chromatin organization of the proximal regulatory domain. J Biol Chem. 2004;279(42):43581–43588. doi: 10.1074/jbc.M408335200. 15. [DOI] [PubMed] [Google Scholar]
- 5.Albergamo A, Apprato G, Silvagno F. The role of vitamin D in supporting health in the COVID-19 era. Int J Mol Sci. 2022;23(7):3621. doi: 10.3390/ijms23073621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Ashley B, Simner C, Manousopoulou A, et al. Placental uptake and metabolism of 25(OH)vitamin D determine its activity within the fetoplacental unit. Elife. 2022;11:e71094. doi: 10.7554/eLife.71094. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Verlinden L, Carmeliet G. Integrated view on the role of vitamin D actions on bone and growth plate homeostasis. JBMR Plus. 2021;5(12):e10577. doi: 10.1002/jbm4.10577. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Roth M, Sun Q, Tamm M. Up-regulated vitamin D receptor by Pelargonium sidoides extract EPs 7630 contributes to rhinovirus defence in bronchial epithelial cells. Pharmaceuticals (Basel) 2021;14(2):172. doi: 10.3390/ph14020172. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Botelho J, Machado V, Proença L, Delgado AS, Mendes JJ. Vitamin D deficiency and oral health: a comprehensive review. Nutrients. 2020;12(5):1471. doi: 10.3390/nu12051471. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dragonas P, El-Sioufi I, Bobetsis YA, Madianos PN. Association of vitamin D with periodontal disease: a narrative review. Oral Health Prev Dent. 2020;;18(1):103–114. doi: 10.3290/j.ohpd.a44323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Antonoglou GN, Suominen AL, Knuuttila M, et al. Associations between serum 25-hydroxyvitamin D and periodontal pocketing and gingival bleeding: results of a study in a non-smoking population in Finland. J Periodontol. 2015;86(6):755–765. doi: 10.1902/jop.2015.140262. [DOI] [PubMed] [Google Scholar]
- 12.Glendenning P, Inderjeeth CA Controversy and consensus regarding vitamin D: recent methodological changes and the risks and benefits of vitamin D supplementation. Crit Rev Clin Lab Sci. 2016;53(1):13–28. doi: 10.3109/10408363.2015.1074157. [DOI] [PubMed] [Google Scholar]
- 13.Schleicher RL, Sternberg MR, Potischman N, et al. Supplemental vitamin D increased serum total 25-hydroxyvitamin D in the US adult population during 2007-2014. J Nutr. 2021;151(8):2446–2454. doi: 10.1093/jn/nxab147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wang K, Xia F, Li Q, Luo X, Wu J. The associations of weekend warrior activity patterns with the visceral adiposity index in us adults: repeated cross-sectional study. JMIR Public Health Surveill. 2023;9:e41973. doi: 10.2196/41973. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Xia F, Li Q, Luo X, Wu J. Association between urinary metals and leukocyte telomere length involving an artificial neural network prediction: findings based on NHANES 1999-2002. Front Public Health. 2022;10 doi: 10.3389/fpubh.2022.963138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xia F, Li Q, Luo X, Wu J. Machine learning model for depression based on heavy metals among aging people: a study with National Health and Nutrition Examination Survey 2017-2018. Front Public Health. 2022;10 doi: 10.3389/fpubh.2022.939758. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Xia F, Li Q, Luo X, Wu J. Identification for heavy metals exposure on osteoarthritis among aging people and machine learning for prediction: a study based on NHANES 2011-2020. Front Public Health. 2022;10 doi: 10.3389/fpubh.2022.906774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Eke PI, Dye BA, Wei L, et al. Prevalence of periodontitis in adults in the United States: 2009 and 2010. J Dent Res. 2012;91(10):914–920. doi: 10.1177/0022034512457373. [DOI] [PubMed] [Google Scholar]
- 19.van de Sande D, van Genderen ME, Verhoef C, et al. Optimizing discharge after major surgery using an artificial intelligence–based decision support tool (DESIRE): an external validation study. Surgery. 2022;172(2):663–669. doi: 10.1016/j.surg.2022.03.031. [DOI] [PubMed] [Google Scholar]
- 20.Zheng X, Shi J, Wu J. Analysis of factors and corresponding interactions influencing clinical management assistant ability using competency model in China. Medicine (Baltimore) 2020;99(51):e23516. doi: 10.1097/MD.0000000000023516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Li W, Zheng Q, Xu M, Zeng C, Deng X. Association between circulating 25-hydroxyvitamin D metabolites and periodontitis: results from the NHANES 2009-2012 and Mendelian randomization study. J Clin Periodontol. 2023;50(2):252–264. doi: 10.1111/jcpe.13736. [DOI] [PubMed] [Google Scholar]
- 22.Acipinar S, Hendek MK, Olgun E, Kisa U. Evaluation of FGF-23 and 25(OH)D3 levels in peri-implant sulcus fluid in peri-implant health and diseases. Clin Implant Dent Relat Res. 2019;21(5):1106–1112. doi: 10.1111/cid.12832. [DOI] [PubMed] [Google Scholar]
- 23.Wierzbicka A, Oczkowicz M. Sex differences in vitamin D metabolism, serum levels and action. Br J Nutr. 2022;128(11):2115–2130. doi: 10.1017/S0007114522000149. [DOI] [PubMed] [Google Scholar]
- 24.Pinto JPNS, Goergen J, Muniz FWMG, Haas AN. Vitamin D levels and risk for periodontal disease: a systematic review. J Periodontal Res. 2018;53(3):298–305. doi: 10.1111/jre.12531. [DOI] [PubMed] [Google Scholar]
- 25.Lu EMC. The role of vitamin D in periodontal health and disease. J Periodontal Res. 2023;58(2):213–224. doi: 10.1111/jre.13083. [DOI] [PubMed] [Google Scholar]
- 26.Arshad S, Zaidi SJA. Vitamin D levels among children, adolescents, adults, and elders in Pakistani population: a cross-sectional study. BMC Public Health. 2022;22(1) doi: 10.1186/s12889-022-14526-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Laky M, Bertl K, Haririan H, et al. Serum levels of 25-hydroxyvitamin D are associated with periodontal disease. Clin Oral Investig. 2017;21(5):1553–1558. doi: 10.1007/s00784-016-1965-2. [DOI] [PubMed] [Google Scholar]
- 28.Millen AE, Hovey KM, LaMonte MJ, et al. Plasma 25-hydroxyvitamin D concentrations and periodontal disease in postmenopausal women. J Periodontol. 2013;84(9):1243–1256. doi: 10.1902/jop.2012.120445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Gümüş P, Öztürk VÖ, Bozkurt E, Emingil G, et al. Evaluation of the gingival inflammation in pregnancy and postpartum via 25-hydroxy-vitamin D3, prostaglandin E 2 and TNF-α levels in saliva. Arch Oral Biol. 2016;63:1–6. doi: 10.1016/j.archoralbio.2015.11.018. [DOI] [PubMed] [Google Scholar]
- 30.Millen AE, Andrews CA, LaMonte MJ, et al. Vitamin D status and 5-year changes in periodontal disease measures among postmenopausal women: the Buffalo OsteoPerio Study. J Periodontol. 2014;85(10):1321–1332. doi: 10.1902/jop.2014.130686. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Hu Z, Zhou F, Xu H. Circulating vitamin C and D concentrations and risk of dental caries and periodontitis: a Mendelian randomization study. J Clin Periodontol. 2022;49(4):335–344. doi: 10.1111/jcpe.13598. [DOI] [PubMed] [Google Scholar]
- 32.Xiao R, Ding C, Hu X, et al. Integrating multimodal information in machine learning for classifying acute myocardial infarction. Physiol Meas. 2023;44(4) doi: 10.1088/1361-6579/acc77f. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.