Research Studies
Study Tracker
Estimating the causal effect of fluoride intake on caries trajectories from childhood to early adulthood with marginal structural models.Abstract
Original abstract online at
https://www.sciencedirect.com/science/article/abs/pii/S030057122500702X?via%3Dihub
Objectives
To identify the longitudinal changes in caries trajectory groups membership from childhood to young adulthood using unsupervised machine learning and estimate the effect of fluoride intake on trajectory group membership using causal inference methods.
Methods
This longitudinal analysis used data from 560 participants in the Iowa Fluoride Study who completed at least two dental examinations at ages 9, 13, 17, and 23. The primary outcome was caries experience measured by D2+MFS counts, and the main exposure variable was total daily fluoride intake from all sources. Our analysis was performed in two stages: 1) Trajectory analysis using the K-means for Longitudinal Data algorithm to identify caries trajectory group membership, 2) Causal inference estimation using marginal structural models with inverse probability weighting using covariate-balancing propensity scores to estimate the effect of fluoride on trajectory membership.
Results
Two trajectory groups were identified: Low (80.2 %) and High (19.8 %) caries trajectory groups. Mean D2+MFS scores rose steadily with age, with the most pronounced increase between ages 13 and 17 in the high trajectory group. Fluoride intake at age 13 demonstrated a significant protective effect (OR=0.15; 95 % CI: 0.04–0.60), with reduced odds of High caries trajectory group membership. The effect of cumulative average fluoride intake was protective but not statistically significant (OR=0.39; 95 % CI: 0.10–1.57). Sensitivity analyses using alternative weight truncation supported the robustness of these estimates.
Conclusion
Fluoride intake during early adolescence appears to be a critical protective factor against high caries trajectory development, highlighting the importance of maintaining adequate fluoride exposure during key developmental periods.
Clinical significance
Adequate fluoride intake during adolescence significantly reduces long-term caries risk by altering disease trajectory patterns. This study identifies early adolescence as a critical window for preventive intervention, supporting targeted fluoride exposure strategies and emphasizing the life-course importance of fluoride in promoting durable enamel health and preventing cumulative oral disease.
Introduction
Estimating the Causal Effect of Fluoride Intake on Caries Trajectories from Childhood to Early Adulthood with Marginal Structural Models
Dental caries is a chronic, cumulative disease that develops over time and can occur at any life stage, particularly when susceptible tooth surfaces are present [1]. Throughout the life course, a range of biological, behavioral, and environmental factors influence oral health, underscoring the importance of adopting a life-course perspective in caries research. Fluoride intake, from both dietary sources and drinking water, remains a cornerstone of caries prevention [[2], [3], [4]].
Despite its significance, relatively few long-term prospective studies have examined dental caries trajectories and explored the causal effect of fluoride intake on trajectory group membership, largely due to the challenges of collecting long-term data and applying appropriate statistical methods [4,5]. Broadbent et al. [6] conducted a study to assess the trajectory of dental caries in a New Zealand birth cohort followed from age 5 to 32. Using a group-based trajectory modeling approach with a zero-inflated Poisson model, the authors identified three distinct trajectories, with higher-risk groups exhibiting nonlinear patterns and substantially greater cumulative caries by adulthood. Similarly, Warren et al. [7]. analyzed longitudinal data from the Iowa Fluoride Study (IFS) using Ward’s hierarchical clustering to define three trajectory groups based on caries experience at ages 9, 13, and 17.
Trajectory modeling methods enable researchers to reduce complex longitudinal data into a few interpretable patterns of disease progression [8]. By grouping individuals with similar developmental trends, these approaches offer a practical framework for identifying populations at varying risk levels [9,10]. In this study, we implemented K-means for Longitudinal Data (KmL), an unsupervised machine learning technique specifically designed to detect latent clusters in longitudinal datasets, even in the presence of missing or irregular data [[11], [12], [13], [14]].
In addition to identifying patterns of caries progression, understanding the modifiable exposures associated with these patterns is critical for prevention. Fluoride intake, from both dietary sources and drinking water, remains a cornerstone of caries prevention [[3], [4], [5], [15], [16]]. However, few studies have rigorously examined the effect of fluoride intake across childhood and adolescence on caries development over time [17,18]. Moreover, even fewer have attempted to estimate the causal effect of total fluoride intake on dental caries trajectory using marginal structural models (MSMs) with inverse probability weighting (IPTW). This causal inference method uses propensity scores to create a pseudo-population in which exposure is independent of measured confounders, enabling estimation of marginal effects [[19], [20], [21]].
This study aimed to investigate the causal effect of long-term fluoride intake on dental caries trajectory using data from the Iowa Fluoride Study. Our analysis proceeded in two stages. First, we identified distinct longitudinal trajectories of dental caries from ages 9 to 23 using KmL. Second, we estimated the effect of total daily fluoride intake on trajectory group membership using MSMs with inverse probability weighting and covariate balancing propensity scores (CBPS) to optimize confounder balance [[19], [20], [21]] This approach provides a less biased estimate of the exposure-outcome relationship by simultaneously addressing time-varying confounding and selection bias, which are paramount challenges in life-course oral health research.
By combining unsupervised learning with causal inference, this study advances the methodological toolkit for longitudinal dental research and provides new evidence on how fluoride intake affects caries development over time.
Section snippets
Study design and population
The Iowa Fluoride Study (IFS) is a prospective cohort that recruited participants from eight Iowa hospitals between March 1992 and February 1995 [22]. Participants were followed with biannual questionnaires and dental examinations at ages 9, 13, 17, and 23, with follow-up concluding in February 2019. The University of Iowa Institutional Review Board approved all study procedures. Informed consent was obtained from participants aged 17 and 23, with parental consent provided for minors. All
Results
The number of IFS participants who completed dental examinations at ages 9, 13, 17, and 23 was 629, 550, 444, and 334, respectively. Among the 560 participants included in the trajectory analysis (those with data from at least two examinations), 52.0 % were female and 48.0 % were male. Concerning family income, 14.3 % reported annual earnings below $40,000, 18.1 % between $40,000 and $59,999, 20.2 % between $60,000 and $79,999, and 47.5 % reported $80,000 or more. Maternal education levels were
Discussion
This study used unsupervised machine learning to identify dental caries trajectories from childhood to early adulthood and employed Marginal Structural Models (MSMs) to estimate the effect of total fluoride intake on caries trajectory group membership.
Of the 560 participants who met the inclusion criteria for trajectory analysis, two distinct caries trajectory groups were identified as optimal based on the Cali?ski-Harabasz criterion. This two-group structure, while differing in number,
Conclusion
Our unsupervised machine learning model revealed two distinct trajectories of cavitated-level dental caries from childhood to early adulthood, with the steepest increases occurring between ages 13 and 17 in the high-risk group. Higher fluoride intake at age 13 was causally linked to a substantially lower risk of belonging to the high caries trajectory group, underscoring fluoride’s protective role during a critical developmental window. These findings support targeted preventive interventions
Ethics approval and consent to participate
Approval for the Iowa Fluoride Study was obtained from the University of Iowa Institutional Review Board. Informed consent (and assent when appropriate) was obtained from all participants and their parents or guardians. All study procedures were conducted in accordance with the Declaration of Helsinki.
Consent for publication
Not applicable
Availability of data and materials
The data supporting the findings of this study are available from the corresponding author upon reasonable request. Due to privacy and ethical considerations, these data are not publicly accessible. A substantial portion of the original data from the Iowa Fluoride Study and the Iowa Bone Development Study was curated and prepared for public sharing via dbGaP as part of NIH grant U01-DE028522, which concluded in 2023.
Funding
This research was funded by the National Institute of Health (grant numbers DE09551, DE12101, RR00059, RR024979), the Roy J. Carver Charitable Trust, the Delta Dental of Iowa Foundation, the Wefel award from the University of Iowa College of Dentistry, and the Post-Comprehensive Graduate Research award from the University of Iowa Graduate College.
CRediT authorship contribution statement
Chukwuebuka Ogwo: Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Philips Okeagu: Formal analysis, Methodology, Writing – original draft, Writing – review & editing. Grant Brown: Formal analysis, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing. Daniel Caplan: Formal analysis, Resources, Supervision, Writing – original
Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Chukwuebuka Elozona Ogwo reports financial support was provided by National Institute of Dental and Craniofacial Research. Chukwuebuka Elozona Ogwo reports financial support was provided by Roy J Carver Charitable Trust. Chukwuebuka Elozona Ogwo reports financial support was provided by The Delta Dental of Iowa Foundation. Chukwuebuka Elozona Ogwo reports
Acknowledgments
We would like to thank Alex Curtis and Chandler Pendelton for their valuable assistance with data management and statistical support.
References (38)
- et al. Sugars and dental caries. Am. J. Clin. Nutr. (2003)
- et al. KmL: a package to cluster longitudinal data. Comput. Methods Programs Biomed. (2011)
- The science and practice of caries prevention. J. Am. Dent. Assoc. (2000)
- The science and practice of caries prevention. J. Am. Dent. Assoc. (2000)
- et al. Oral diseases: a global public health challenge. Lancet (2019)
- World Health Organization, Oral Health Fact Sheet, World Health Organization, Geneva,…
-
Recommendations for using fluoride to prevent and control dental caries in the United States. MMWR. Recomm. Rep. (2001)
-
Water fluoridation review. Cochrane Database Syst. Rev. (2015)
- et al. Applying modern survival analysis methods to longitudinal dental caries studies. J. Dent. Res. (2002)
- et al. Trajectory patterns of dental caries experience in the permanent dentition to the fourth decade of life. J. Dent. Res. (2008)
- et al. Dental caries clusters among adolescents. Community Dent. Oral Epidemiol. (2017)
- Kaufman L., Rousseeuw P.J., Finding Groups in Data: An Introduction to Cluster Analysis, John Wiley & Sons, New York,…
- Nagin D.S., Group-Based Modelling of Development, Harvard University Press, Cambridge, MA,…
- et al. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol. Methods Res. (2007)
- et al. KmL: k-means for longitudinal data. Comput. Stat. (2010)
- et al. Identifying typical trajectories in longitudinal data: modelling strategies and interpretations. Eur. J. Epidemiol. (2020)
- et al. Algorithm AS 136: a K-means clustering algorithm. J R Stat Soc C Appl Stat. (1979)
- Public Health England, Water Fluoridation: Health Monitoring Report for England 2018, Public Health England, London,…
- et al. Considerations on optimal fluoride intake using dental fluorosis and dental caries outcomes—A longitudinal study. J. Public Health Dent. (2009)