Review and recommendations for zero-inflated count regression modeling of dental caries indices in epidemiological studies.

Over the past 5-10 years, zero-inflated (ZI) count regression models have been increasingly applied to the analysis of dental caries indices (e.g. DMFT, dfms). The main reason for that is linked to the broad decline in children's caries experience, such that dmf and DMF indices more frequently generate low or even zero counts. This article specifically reviews the application of ZI Poisson and ZI negative binomial regression models to dental caries, with emphasis on the description of the models and the interpretation of fitted model results given the study goals. The review finds that interpretations provided in the published caries research are often imprecise or inadvertently misleading, particularly with respect to failing to discriminate between inference for the class of susceptible persons defined by such models and inference for the sampled population in terms of overall exposure effects. Recommendations are provided to enhance the use as well as the interpretation and reporting of results of count regression models when applied to epidemiological studies of dental caries.

The classic caries clinical trial: constraints and opportunities.

The history of clinical trials would include events in 1747 on board the Salisbury, a British Navy vessel at sea with 12 seamen critically ill with scurvy. Involving these 12 sailors in a study, an officer on board by the name of Lind evaluated six potential treatments for scurvy, and rapidly reached the conclusion that daily consumption of citrus fruits returned the men fit for duty in approximately six days (Bull, 1959). The concept of experimental randomization was first developed by Sir R.A. Fisher (1925, 1926), and the method was introduced to medical research via a study of tuberculosis treatment by Amberson and co-workers (1931), who randomized 24 TB patients into two groups, one to receive the experimental therapy, the other serving as the control. Amberson et al. also incorporated the concept of blinding into their study. Sir Austin Bradford Hill codified and built on the principles of scientific experimentation developed by Fisher, and introduced the use of random numbers in the allocation of patients in the British Medical Research Council (1948) study of the effect of streptomycin in the treatment of tuberculosis (Daniels and Hill, 1952; Hill, 1952). The first applications of clinical trial methodology for testing interventions on dental, oral, and maxillofacial diseases and conditions are more difficult to determine. For dental caries prevention, however, Chilton and Fertig (1958) and Slack and Martin (1964) were certainly among the early caries clinical trial pioneers. As clinical trials have come into the mainstream of clinical research in medicine and dentistry, a great deal of developmental work has focused on their methodological enhancement. The most successful of these efforts have come from fruitful, ongoing collaborations among clinician investigators, biostatisticians, data management specialists, biomedical ethicists, and others with an academic interest in clinical trial design and utilization. During the past 25 years, the emergence of systematic reviews and the evidence-based medicine (EBM) movement have also contributed significantly to the increasing reliance on randomized clinical trial outcomes for the advancement of better clinical practice (Richards et al., 1997; Straus and Sackett, 1998; www.cochrane.org/cochrane/ccbroch.htm#BDL, 2002).

Types of clinical caries studies: epidemiological surveys, randomized clinical trials, and demonstration programs.

This has been a necessarily cursory overview of three types of clinical dental caries studies: the dental survey, the randomized clinical trial, and the non-randomized comparative study. Each of these studies may be used in its appropriate place, although only the survey and the randomized clinical trial are well-grounded in scientific theory. In general, the dental survey is well-suited to gathering basic cross-sectional data on dental caries, data that serve a variety of important practical and scientific interests in the dental care system. Clinical trials and non-randomized studies are applied in evaluating a variety of therapies for dental caries. Of the two, the randomized clinical trial is by far the more powerful. As a result, the clinical trial has found a central place in dentistry. The contributions that past and present clinical caries trials have made to dental health are quite evident to most observers. It is virtually inescapable that well-conducted clinical caries trials will make even more significant contributions to dental practice in the future, thereby further improving the oral health of all.

Classification tree prediction models for dental caries from clinical, microbiological, and interview data.

Caries prediction by Classification And Regression Tree (CART) analysis is an appropriate and powerful alternative or complement to the commonly used classification methods of logistic regression and discriminant analysis, both parametric and nonparametric. The binary classification tree method discussed in this article is designed for complex data and does not require assumptions about the predictor variables or about the presence or absence of interactions among the predictor variables. Furthermore, the results give insight into the structures and interactions in the data and are easy to interpret and apply. In preliminary applications of the CART algorithms to data from The University of North Carolina Caries Risk Assessment Study, the method produced prediction rules having sensitivities and specificities that were similar to or slightly better than those associated with logistic and discriminant analyses. The classification trees constructed tended to involve far fewer predictor variables than required for adequate logistic and discriminant models. For example, for first-grade children in Aiken, South Carolina, nine variables were used to define a prediction rule having 64% sensitivity and 86% specificity. Ten-fold cross-validation estimates for future data were 58% and 79%, respectively. For first-grade children in Portland, Maine, two variables were used to define a prediction rule having 62% sensitivity and 77% specificity. The cross-validation estimates for future data were 58% and 78%, respectively. A brief, and previously unavailable, explanation of the CART method is given for the special case of a dichotomous outcome variable.

Relationship of total fluoride intake to beneficial effects and enamel fluorosis.

Recent studies indicate that the prevalence of very mild to moderate dental fluorosis, as classified by Dean, has increased relative to that found in earlier investigations. To date, fluoridated water, fluoride supplements, the diet, fluoride dentifrices, and other topical fluoride applications have been identified as sources of systemic fluoride. Recent evidence suggests that there is a strong association between mild to moderate enamel fluorosis and the use of fluoride supplements during early childhood, and that the presently recommended supplementation schedule for U.S. children above the age of 2 years may be too high. Evidence also suggests that there is a strong association between fluoride dentifrice use during early childhood and enamel fluorosis in fluoridated populations. These findings support the need for a careful review of existing supplementation schedules and early oral hygiene practices. There is a pressing need for additional analytical epidemiological studies to confirm existing findings and to determine whether other fluoride sources may be associated with enamel fluorosis. Further, since exposure to combinations of individual risk factors has been shown to carry more than merely an additive increase in the risk of fluorosis, these studies must be multifactorial in design. There is also a need for more fluorosis prevalence and severity data to be gathered, so that the development of enamel fluorosis as a public health problem can be assessed, and so that the success of measures implemented to maximize efficacy while minimizing unwanted side-effects can be monitored.

Surgical removal of the fully impacted mandibular third molar. The influence of flap design and alveolar bone height on the periodontal status of the second molar.

This study was carried out because a great deal of uncertainty exists as to he effect of the surgical removal of the fully impacted third molar on the periodontal status of the second molar. The objectives were to determine the effect of surgical removal of the third molar on the periodontal status of the second molar; the influence of flap design on these results; the influence of the initial height of the alveolar bone on the distal of the second molar on subsequent changes in attachment level. The study included 30 patients with bilateral mandibular impactions. A split-mouth experimental design was used, with one side of the mandible being randomly allocated to one of two flap design groups. Plaque level, gingival inflammation, probing depth and attachment level measurements around the second molar were taken at baseline and then at monthly intervals for a period of 6 months. Alveolar bone height was measured from panoramic radiographs. Six months postsurgically, both flap design groups exhibited a statistically significant loss of attachment level on the distal surface of the second molar with no difference between the two flap groups. The initial height of the alveolar bone on the distal of the second molar had no influence on the loss of attachment. It was concluded that the surgical removal of the fully impacted mandibular third molar led to the loss of attachment on the distal of the second molar; flap design had no influence on the degree of attachment loss; the initial height of the alveolar bone on the distal of the second molar had no influence on the loss of attachment.

Comparative effectiveness of flossing and brushing in reducing interproximal bleeding.

This study assesses the comparative effectiveness of three types of dental floss and toothbrushing in reducing interproximal bleeding sites, measured by stimulation with wooden interdental cleaners using the interdental bleeding index assessment method. The 119 adult subjects with gingival inflammation were randomly assigned to one of four groups at the beginning of a supervised 2 week clinical trial. The toothbrushing only group achieved a 35% reduction in bleeding sites and the three flossing groups all demonstrated dramatic reductions of about 67%. The three varieties of dental flosses were about equally effective in reducing interproximal bleeding and doubly effective as toothbrushing alone. Compared to probing, the interdental bleeding index method is considered a simplified method of assessing interproximal gingival inflammation.

Influence of a diet of fluoride-fed cockerels on reproductive performance of captive American kestrels.

In 1982, twenty-four pairs of captive American kestrels (Falco sparverius) were forced to renest by removal of their first clutches 6 days after their completion. Immediately following, each of three groups of eight pairs was randomly assigned to one of three daily dietary regimes for 10 days: (1) three 1-day old cockerels with background levels of F(-) (62.4+/-51ppm, mean+/-SD) in their femurae, (2) two 10-day old cockerels with 4512+/-810ppm of F(-) in their femurae, (3) two 10-day old cockerels with 7690+/-417ppm of F(-) in their femurae. Fluoride levels in femurae of treated kestrels were significantly (P<0.0025) higher than those of control birds. Clutch sizes tended to be smaller as more fluoride was added to the diet, but not significantly so, due to an increase of the variance in the treatment group. Per cent fertility and per cent hatchability were not significantly affected by treatment. The fluoride content in eggshells in the fluoride-treated groups differed significantly from those of the control group (P<0.001).

Results of the Sherbrooke-Lac Mégantic fluoride varnish study after 20 months.

The purpose of this study was to demonstrate the effectiveness of two fluoride varnishes (Duraphat and Fluor Protector) in the prevention of dental caries in children residing in a non-fluoridated community. At baseline, 787 grade one children (6-7-yr old) were examined and randomly allocated into a control and two treatment groups. At 20 months, 703 children remained in the study. The mean baseline caries experience and the mean number of permanent teeth present at 20 months were not significantly different among the treatment and control groups. The mean DMFS increments after 20 months were 1.70, 1.73 and 2.02 for Fluor-Protector, Duraphat and control groups respectively (% reductions of 15.8 and 14.4 for Fluor-Protector and Duraphat groups). The mean dfs increments for primary molars after 20 months were 1.56, 1.62 and 1.74 for Fluor-Protector, Duraphat and control groups respectively (% reductions of 10 for Fluor-Protector and 7 for Duraphat).

Comparative effects of a 4-year fluoride mouthrinse program on high and low caries forming grade 1 children.

This paper presents the comparative effectiveness of fluoride mouthrinse (FMR) on high and low caries forming children after a 4-yr exposure to weekly rinse beginning in the first grade. Over 1200 grade 1 children drawn from both fluoride deficient and fluoridated sites were divided into treatment and concurrent, longitudinal control groups. After 4 yr these children were stratified according to caries increment; those above the 75th percentile were considered high caries formers, all others were designated low caries formers. After adjustment of the mean increments for differences in SES, age, race, and sex in rinse and control groups, high caries formers (approximately 25% of the children) in the rinse and control groups in fluoride deficient areas showed increments of 7.00 and 7.79 surfaces, respectively, indicating a savings of 0.79 surfaces. Low caries formers (approximately 75% of the children) demonstrated increments of 1.11 DMFS in the rinse group and 1.40 in the control group (savings 0.29 DMFS). The pattern was quite similar for children in fluoridated areas except that the increments, as well as the savings realized, were lower. The results raise questions as to the practical effectiveness of school based FMR programs even for high caries forming children.

University of North Carolina Caries Risk Assessment Study: comparisons of high risk prediction, any risk prediction, and any risk etiologic models.

The purpose of this analysis is to compare three different statistical models for predicting children likely to be at risk of developing dental caries over a 3-yr period. Data are based on 4117 children who participated in the University of North Carolina Caries Risk Assessment Study, a longitudinal study conducted in the Aiken, South Carolina, and Portland, Maine areas. The three models differed with respect to either the types of variables included or the definition of disease outcome. The two "Prediction" models included both risk factor variables thought to cause dental caries and indicator variables that are associated with dental caries, but are not thought to be causal for the disease. The "Etiologic" model included only etiologic factors as variables. A dichotomous outcome measure--none or any 3-yr increment, was used in the "Any Risk Etiologic model" and the "Any Risk Prediction Model". Another outcome, based on a gradient measure of disease, was used in the "High Risk Prediction Model". The variables that are significant in these models vary across grades and sites, but are more consistent among the Etiologic model than the Predictor models. However, among the three sets of models, the Any Risk Prediction Models have the highest sensitivity and positive predictive values, whereas the High Risk Prediction Models have the highest specificity and negative predictive values. Considerations in determining model preference are discussed.

Identifying children who will experience high caries increments.

An 18-month longitudinal study of 197 5 to 18-yr-old children was conducted to assess the performance of multiple baseline variables in predicting which children would experience high increments of caries. Nine predictors were assessed; DMFS and defs scores, number of permanent teeth, indices of fissure retentiveness and occlusal morphology, S. mutans and lactobacilli levels, and sex and race. Discriminant analyses were applied to data for children in five age groups to identify the 30% who would have the largest 18-month DMFS increments. Prediction performance was assessed by comparisons with the actual high increment group, defined as those children with increments in the upper quartiles of the DMFS distributions within age groups. The analyses predicted between 56% and 91% of actual high increment children depending on age group. The children identified in the analyses experienced between 59% and 91% of the disease experienced by the actual high increment groups.

Development and application of a prediction model for dental caries.

The development and validation of a caries prediction model comprising 13 sociodemographic and dental examination variables on Grade 1 and Grade 5 children in the National Preventive Dentistry Demonstration Program are described. The objective was to derive a method of predicting children at high risk to caries early in order that preventive measures might be undertaken. True high risk children were defined in two ways: highest 25% of children based on their 4-yr DMFS increment, and their total DMFS score at the end of the study. In both cases, children predicted to be at high risk were defined as the 25% with the highest discriminant score. Discriminant function and logistic regression analyses were used to determine the extent to which the 13 variables collectively discriminated between true high risk and non-high risk children so defined. Sensitivity was approximately 0.50 and specificity around 0.82, using the 4-yr increment as the criterion for defining true high risk, and approximately 0.64 and 0.88, respectively, using the final DMFS score for defining true high risk.

The University of North Carolina caries risk assessment study: caries increments of misclassified children.

The University of North Carolina caries risk assessment was conducted between 1986 and 1989 with 5000 children initially in grades 1 and 5 from low fluoride sites in South Carolina and Maine. Clinical, microbiologic, behavioral, and demographic factors served as independent variables used in logistic multiple regression models to determine the predicted caries risk classification of each child. The 3-yr DMFS increment of each child was the dependent variable, and the 20-25% of those in each cohort with the highest increment were considered the high risk group. Sensitivity and specificity values averaged 0.60 and 0.83 respectively, indicating the misclassification of substantial numbers of children. This paper reports analysis of the DMFS increments of the misclassified children. The majority of children had DMFS increments within one or two surfaces of the classification cut points. Many false negative children had increments composed entirely of filled surfaces, often of minimal extent and without evidence of previous decay. Many false positive children had increments composed of decayed surfaces and would have benefitted from being identified as high risk and assigned to a preventive program. Other false positive children had sealants placed after the baseline examination that undoubtedly reduced their true increment. It is concluded that the consequences of misclassification are not serious for most children in this study.

The University of North Carolina Caries Risk Assessment study: further developments in caries risk prediction.

Over 4000 first and fifth grade children from the areas surrounding Aiken, South Carolina, and Portland, Maine, participated in a 4-yr study to develop caries risk assessment models. The predictors used at baseline included detailed clinical examinations, salivary microbiological tests, and sociodemographic and dental behavior data. Mean 3-yr caries increments in South Carolina were twice those in Maine. For the four risk assessment models (two grade cohorts at two sites) specificity values averaged 0.83 and sensitivity values averaged 0.60. Clinical predictors such as prior DMFS, pit and fissure morphology, and predicted caries risk status were the major contributors to the models.

The use of fissure sealants in public health programs: a reactor's comments.

The papers presented a thorough, up-to-the-minute look at fissure sealants and their possible use in dental public health programs. The papers demonstrated that sealant technology has progressed to the point that where properly used, retention rates and caries-prevention effectiveness are, even over a five-year period, satisfactory. As a reactor to the papers, I attempted to underline three points: First, with respect to the sealant-amalgam comparison, the point must be made forcefully that sealants are not an alternative to amalgams. Rather, they represent a highly effective and technically practical procedure that, if correctly employed in a preventive program, would significantly reduce the cutting of teeth and the placement of amalgams. This is particularly relevant to the use of sealants in private practice. Second, in formal public health programs, the use of sealants should probably be concentrated on the more highly caries-susceptible children in the population who would require a full range of chairside preventive procedures. Third, for the immediate future there is a pressing need for a limited number of well-defined studies to further investigate the cost and price determinants associated with the fissure-sealant procedure. These points were all explicitly or implicitly made in the papers presented and I appreciate having had the opportunity to reemphasize them.

An epidemiologic estimate of the critical period during which human maxillary central incisors are most susceptible to fluorosis.

The temporal relation between a declining fluorosis gradient and an abrupt downward shift in community drinking water fluoride concentration was evaluated through multiple correlation analysis to determine the critical time frame during which developing maxillary central incisors are most susceptible to fluoride challenge. Fluorosis data were scrutinized through a time-related series of epidemiologic "windows" or time frames of varying lengths. The placement of these time frames was in turn related to the presumed start of enamel mineralization (at birth), and ranged from zero to 60 months later. In this way, the susceptibility of developing enamel to changes in water fluoride concentration was localized. The greatest risk was associated with a four-month critical period commencing at 22 months following birth. The risk of fluorosis from exposures to a fluoride challenge acting during shorter periods was better localized than risk associated with longer exposures. We concluded (1) that human maxillary central incisors are most susceptible to fluorosis during a critical period of as little as four months' duration, commencing at 22 months of age; and (2) that for these incisors, fluoride exposure during the months prior to this period carries less risk than continued exposure for up to 36 months beyond this critical time.

Dental fluorosis following downward adjustment of fluoride in drinking water.

The drinking water fluoride concentration in Hong Kong was reduced by about 0.2 ppm in June 1978. This study was undertaken to determine whether the prevailing level of dental fluorosis was affected by such a minor change. Cohorts of children (N = 1,062) aged seven to 12 years, who were born both before and after the fluoride reduction, were examined clinically using Dean's fluorosis index. Based on upper right central incisors, dental fluorosis prevalence decreased from 64 to 47 percent and the community fluorosis index decreased from 1.01 to 0.75 (P less than .01). Thus, dental fluorosis was reduced, although the reduction in water fluoride concentration was not sufficient to achieve the minimal fluorosis level that Dean associated with a fluoride concentration optimal for caries prevention. Variation in dental fluorosis has been reported previously to result from marked sudden changes, during tooth formative years, to drinking water fluoride concentration. This study confirms preliminary findings that variation in dental fluorosis arising from minor changes to the fluoride level in drinking water is also measurable.

University of North Carolina caries risk assessment study. III. Multiple factors in caries prevalence.

The baseline caries experienced of approximately 5,000 children in South Carolina and Maine was used as the dependent variable in caries risk assessment analyses. Clinical, microbiologic, and demographic factors served as independent variables in a multivariate relationship to caries through regression and discriminant function analyses. Four factors--number of dental visits by the child in the past year, presence of white spot lesions, and both the urgency of need for restorative care and the future caries increment predicted by the examiner--associated significantly and consistently with caries prevalence in primary and permanent teeth of first and fifth graders at both study sites. Several factors associated significantly with caries prevalence at only one site or grade within a site, suggesting that wide applicability of a specific caries risk assessment model may be limited. In these analyses, sensitivity ranged from .60 to .72 and specificity varied from .86 to .91 in the four grade-site groups. The ultimate goal of this longitudinal study is to identify highly caries-prone children in time to prevent the occurrence of a future caries increment. Although the lack of consistent association of many variables, including microbiologic factors, with baseline caries prevalence was unexpected, it is expected that some of these variables will contribute predictive power in the prospective study.

The University of North Carolina Caries Risk Assessment Study. I: Rationale and content.

This article presents the rationale and content of a current study that seeks to improve methods to identify children at high risk to dental caries. It summarizes the results of the development of a 12-factor, preliminary caries prediction model based on data derived from the National Preventive Demonstration Program. Despite data limitations, the model produced a sensitivity of .5 and specificity of .8 for four-year caries increment prediction in first- and fifth-grade children. Data on a number of additional potential predictors are being collected in two sites to expand and improve the existing model. These factors are identified.