Assessing marginal associations between dental caries and periodontal diseases and their determinants in cross-sectional studies

This proposed research is both for statistical methodology development and novel secondary analyses of an important existing national dataset. The overall goal of our proposed research is to assess person-level marginal associations between dental caries and periodontitis. We consider a modeling setup where individual units (e.g., teeth and surfaces) are clustered so that the outcomes in a given cluster (e.g., mouth) are not independent. Due to the nature of the underlying biology, both caries and periodontal outcomes are correlated with the size of the cluster. This phenomenon is referred to as “informative cluster size (ICS).” Additionally, treating a periodontal outcome such as the attachment loss as one of the two variables in this association study, we may face a greater form of informativeness that goes beyond ICS; e.g., many more teeth with smaller values than larger values of attachment loss, even for subjects with comparable numbers for total cluster size (e.g., number of teeth). As a result, a different type of reweighting will be needed to assess the marginal associations. After developing the general statistical methodology for calculating these novel associations by solving a generalized estimation equation, we will apply a regression technique based on jackknife pseudo-values [Andersen et al, 2003] to determine which of the modeling factors (e.g., sex, race, economic status, family income, smoking status, education level, tooth position, and so on) significantly contribute to this association. A motivating dental dataset to which to apply our statistical research is from the National Health and Nutrition Examination Survey (NHANES). Tooth and surface/site level cross-sectional national US data stratified by age (30-49, 50-64, and 65+) will be analyzed using our novel statistical methods. During the proposed research period of two years, we will undertake the following tasks: Aim 1: We will develop statistical tests based on resampling combined with permutation to test for informativeness of number of observed data units (e.g., teeth) for caries outcomes at various levels of attachment losses. Aim 2: We will introduce novel inferential methods for the marginal associations (with different marginalization weights) between caries and periodontal outcomes and those of various covariate effects and use in detailed analyses of the NHANES data set of periodontal and caries outcomes mentioned above. Aim 3: We will create a scalable, computationally-efficient, and user-friendly R-package based on our inferential tools to be developed under Aims 1 and 2. Without this R-package, members of the scientific community may not be able to our novel methods for their data analysis, since the base R software does not have implementations of these methods. The R-package will be tested on the NHANES data set, which will be supplied with this R-package. Finally, the R package will be distributed to the scientific community at large through CRAN (Comprehensive R Archive Network). Our proposed research is important both for gaining new scientific knowledge and for statistical methodology development leading to new tools of data analysis. Project Number: 1R03DE033774-01A1 | Fiscal Year: 2025 | NIH Institute/Center: National Institute of Dental and Craniofacial Research (NIDCR) | Principal Investigator: Somnath Datta | Institution: UNIVERSITY OF FLORIDA, GAINESVILLE, FL | Award Amount: $319,624 | Activity Code: R03 | Study Section: Special Emphasis Panel[ZDE1 TO (04)] View on NIH RePORTER: https://reporter.nih.gov/project-details/11130505