The Application of Multilevel Modeling in the Analysis of Association between Periodontal Disease and Metabolic Syndrome
Date Issued
2011
Date
2011
Author(s)
Lin, Yu-Jie
Abstract
Abstract
Background: Periodontal diseases (PD), referring to the destruction of tissues surrounding the teeth, can cause teeth loss, and affect chewing function and speaking ability. In the past, most of investigations for periodontal disease focused on subject level. In this study, the periodontal status for subject was measured in 6 sextants, which are highly correlated in the same subject. To enhance the statistical efficiency and copy with correlated property among data of periodontal status on six sextants within the same subject, the multilevel models were used.
Aim: In this study, multilevel models were applied to investigate the association between PD and metabolic syndrome adjusting with other risk factors.
Methods and Materials: Data were from a large-scale periodontal diseases survey project which covered 99 different districts and was conducted between 2007 and 2008. A total of 4061 participants aged 18 or above were enrolled in this study. Periodontal status was measured with CPI and LA. Results of biochemical examination were recorded for each participant. Information of oral hygiene behavior and life style was obtained by interview with structured questionnaire. We used logistic regression to analyze the relationship between periodontal disease and metabolic syndrome. Furthermore, linear regression and generalized liner mixed models are used to correlate metabolic syndrome and the severity of periodontal disease in both CPI and LA.
Results: The overall age-adjusted prevalence rate of PD, defined as CPI≥3, was 52.00%. Among male, the prevalence rate of PD was 56.89% which was higher than that among female, 48.89%. As PD was defined as LA≥1, the overall age-adjusted prevalence rate of PD was 49.33% (57.31% for male male, and 44.09% for female). Prevalence rate increased with age. Waist circumference, blood glucose and cholesterol are significantly associated with LA, but not with CPI. Blood pressure is significantly associated with CPI, but not with LA. Number of criteria for metabolic syndrome met was positively significantly associated with severity with CPI, both in linear regression and mixed models. For other confounders, mixed model was more efficient compared with linear regression model.
Conclusion: Our study showed that results in sextant level with mixed model was more efficient compared to those in subject level with linear regression model when severity of periodontal disease was used. Nevertheless, number of criteria for metabolic syndrome met was positively significantly associated with severity with CPI in both models.
Subjects
Multilevel Modeling
Periodontal Disease
Metabolic Syndrome
random effect
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-100-R97842001-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):592d8a6b088fc9ac62dbc925e5480bd0