Skip to Main Content

NIH: Consensus Development Conference on Diagnosis and Management of Dental Caries Throughout Life: Background

NIH organized conference that produced consensus statements on important and controversial topics in medicine and dentistry.

Clinical Applications and Outcomes of Using Indicators of Risk in Caries Management

Domenick T. Zero, D.D.S., M.S., Margherita Fontana, D.D.S., Ph.D., and Aine M. Lennon, B.Dent.Sc., Ph.D.:

Other papers at this conference have discussed individual risk indicators of caries. This review focuses on studies of the predictive validity of various combinations of risk indicators. Such indicators may be useful in the clinical management of dental caries by helping dental professionals determine if additional diagnostic procedures are required, identify patients who require caries control measures, assess the impact of caries control measures, make treatment planning decisions, and determine the timing of recall appointments. Although there is a high level of interest in identifying risk indicators, only a few studies have attempted to determine how the application of risk indicators affects dental health outcomes (Brambilla, Gagliani, Felloni, et al., 1999; Hausen, Karkkinena, Seppa, et al., 2000).

Multifactorial modeling has proved its value in longitudinal caries prediction studies by showing the interrelations and interactions of risk factors. Beck and colleagues (1988) indicated that one or more social, behavioral, microbiologic, environmental, and clinical variables should be included in such a model, given the many factors that influence dental caries. Modeling has usually been based on a dichotomized dependent variable, either as "no" versus "some" caries increment (Beck, Weintraub, Disney, et al., 1992) or with specified cut-off points in populations with high caries incidence (Abernathy, Graves, Bohannon, et al., 1987). The accuracy of models has rarely been 80 percent, which is considered to be the minimum level for screening purposes. "To be useful, a working model should produce a sensitivity of 0.75 or higher and specificity level of at least 0.85 or higher" (Stamm, Disney, Graves, et al., 1988). It has therefore been suggested that a risk model should have a combined sensitivity and specificity of at least 160 percent (Kingman, 1990).

Objective

The aim of this review was to systematically assess the clinical evidence to determine the predictive validities of currently available multivariate caries risk-assessment strategies. The intent was to answer "What are the best (combination of) indicators for an increased risk of dental caries?" That, in turn, should help to answer Question 5, "How should clinical decisions regarding prevention and/or treatment be affected by detection methods and risk assessment?"

Search Strategy

A search of relevant publications dating from 1980 was conducted in the MEDLINE and EMBASE databases. Only English language publications concerning humans were included in the search. To help identify as many papers as possible the following key word headings were used:

  • For primary dentition: [(Caries AND Risk hedge) AND Diagnosis hedge/limited to human, English, 1980+] AND (age group limit OR primary dentition hedge).

  • For root caries: [(Caries AND Risk hedge) AND Diagnosis hedge/limited to human, English, 1980+] NOT (age group limit OR primary dentition hedge) AND root caries hedge.

  • For permanent dentition: [(Caries AND Risk hedge) AND Diagnosis hedge/limited to human, English, 1980+] NOT [(age group limit OR primary dentition hedge) OR root caries hedge].

Due to the large number of references obtained in our electronic search, it was decided that secondary hand searching would not be feasible.

Selection Criteria

Inclusion and exclusion criteria for the papers selected for review included: (1) the use of more than one type of caries risk predictor category used to calculate the predictive outcome, and (2) the presence of a clear outcome prediction. Every included article was listed, as were excluded articles. The following types of articles were excluded: reviews, in vitro studies, research using population approaches rather than individual approaches, and papers not related to dentistry. Except for review papers, these are not listed in the exclusion table.

Data Collection and Analysis

A list of included and excluded articles for each category (primary teeth, permanent teeth, and root caries) was prepared. At the time of preparation of this abstract, 151 papers had been added to either the inclusion or exclusion tables, and 27 were still being sought. Papers that conformed to the selection criteria and reported a predictive outcome for the model were included (N= 24 for primary teeth; N= 37 for permanent teeth; and N= 13 for root caries). The tabulation of excluded articles (N= 77) included the reason for exclusion (e.g., lack of more than one risk factor, no outcome data, etc). Four evidence tables were prepared: primary teeth, permanent teeth in children and/or adolescents, permanent teeth in adults, and root caries. When an article appeared in one data set (e.g., primary teeth) but contained information on another data set, it was transferred to the appropriate inclusion table. Articles reporting information on more than one type of caries were included in more than one table. Included articles were also grouped by study design as longitudinal-prospective, retrospective, or cross-sectional.

Of the 24 articles on primary teeth, 17 were prospective studies, 1 was a retrospective study, and 6 were cross-sectional studies. The articles on permanent teeth were separated into those involving caries in risk prediction in children/ adolescents (< 20 years old) and those used to predict caries in adults. Of the 30 articles on permanent teeth in children/adolescents, 20 were prospective studies, 2 were retrospective studies, and 8 were cross-sectional studies. Of the total of 7 articles on permanent teeth in adults, 2 were prospective studies and 5 were cross-sectional studies. For root caries, 13 articles were found: 9 prospective studies and 4 cross-sectional studies. All models included some aspect of past caries experience as a predictor. The second most frequent predictor was "other variables." The third most frequent predictor was "microflora," followed by "host factors." In the case of root caries the "host factors" category was more frequently used than the "microbiology" category.

References were systematically assessed for their validity. Since valid evidence is considered best obtained from randomized, controlled longitudinal (prospective) studies, those were given the highest scores in our review. Studies were graded as "good," "fair," or "poor," depending on the amount of information they provided to support the methodology used. The main variables assessed for this purpose (other than the inclusion criteria) were: (1) whether the study reported how samples were obtained, (2) whether the examiners were trained/calibrated, (3) whether examiner reliability was reported, and (4) whether examiners were blinded during the study. Tables 1, 2, and 3 include the longitudinal prospective studies considered to be good sources of evidence for predictions in primary teeth, permanent teeth in children and adolescents, and permanent teeth in adults. None of the root caries studies reviewed met these criteria.

Of all the models reviewed, none of those graded as "good" had a combined sensitivity and specificity in excess of 160 percent, although the model reported by Demers and colleagues (1992) comes very close (159 percent). These authors concluded that previous caries experience was the strongest predictor in their model, followed by parents� education. For primary teeth there was one "fair" study in which combined sensitivities and specificities totaled 170 percent (Holst, Martensson, Lavrin, et al., 1997). That study used infants 1 year old, for 2 years, and all categories of risk assessment factors. Visible plaque, deep fissures, and oral hygiene were the strongest predictors.

Table 1. Primary teeth-prospective studies (good level of evidence)

Researcher

N

Age at onset

Study Design

Variables:
Past Caries
or Disease Experience

Variables: Microflora

Variables:
Host

Variables:
Other

Outcome =
Validation criteria = true disease

Sensitivity

Specificity

[Isokangas et al., 1993]

297 (3-4 year olds)

3-4

Prospective
(1 year)

Caries, Predicted caries

Not used

Not used

Sociodemographic

<1 dentinal caries lesion in need of restoration

(actual data NR)

45%

92%

[Demers et al., 1992]

302

5 year olds

Prospective

(1 year)

Caries experience:
dmfs=0 or dmfs>0 (WHO, no radiographs)

SM, LB (Bactotest)

Buffer capacity

Age, sex, parent’s education, family structure, fluoride consumption, oral hygiene (debris index)

>1 ds

(mean dfs increment:
2.1 + 3.6)

81.8%

78.3%
(for caries experience only)

77.4%

77.4%
(for caries experience only)


* Bold: included in final models or strongest predictors
MS: mutans streptococci
LB: Lactobacilli
LRA: logistic regression analysis
LDA: logistic discriminant analysis
NR: Not reported

Table 1. Primary teeth-prospective studies (good level of evidence) (continued)

Researcher

Baseline Scores
(Mean + SD)

True High Risk Criteria Used

Method of modeling

Country

Sampling method

Training of examiners reported

Reliability of examiners

Blinding of examiners

Blinding of patients

Subject Attrition

Authors conclusion

[Isokangas et al., 1993]

NR

High risk: Any caries increment

Not used

Finland (Ylivieska)

All 3-16 year olds in public dental care were included

15 clinicians participated. No training reported.

NR (dentists examined different children)

Not possible for ethical reasons

NR

NR

Clinicians can predict risk using only caries and socio-demographic variables available at annual examinations

[Demers et al., 1992]

NR

At least one new carious lesion in primary teeth: high risk

(LRA; 9 variables studied)

Canada (Montreal)

Non-fluoridated community

Random selection of schools

Calibrated (2 examiners)

For caries:
Intraexaminer reliability: intraclass correlation coefficient >0.95.
The same true for interexaminer reliability

For micro test:
Intraexaminer reliability:0.80-1.00; interexaminer reliability: 0.79-0.87.

NR

NR

126

Previous caries experience was the best predictor, followed by parent’s education.

Table 2. Permanent teeth-children and adolescents; prospective studies (good level of evidence)

Researcher

N

Age at onset

Study Design

Variables: Past Caries or Disease Experience

Variables: Microflora

Variables:
Host

Variables: Other

Outcome=Validation criteria=true disease

Sensitivity

Specificity

[Disney et al., 1992b]|
North Carolina Study
"High Risk Prediction Model"

4158:

2079 (Aiken, GA)

2096 (Portland, ME)
Both: fluoride deficient, high caries experience

6 years
(1st grade)
and 10 years old (5th grade)

Prospective
(3 years)

DMFS (Radike, no radiographs), dmfs, predicted caries; fluorosis, white spot lesions

SM (Cariescreen), LB (Bactotest), mean plaque score

Pit and Fissure Morphology

Sociodemographic (higher in Portland-exclusively white); examiner, age, brushing frequency, between meals snacks

>4 DMFS



> 2 DMFS


(At 3 years-DMFS increment:

Aiken: 1.9 (grade 1), 3.1 (grade 5)
Portland: 0.8 (grade 1), 1.5 (grade 5)

59% (grade 1); 62%
grade 5

59% (grade 1); 62% (grade 5)

83% (grade 1); 81% (grade 5)

84% (grade 1); 84% (grade 5)

[Isokangas et al., 1993]

1464 (5—16 year olds)

 

3—16

Prospective
(1 year)

Caries, Predicted caries

Not used

Not used

Socio-demographic

<1 dentinal caries lesion in need of restoration

(actual data NR)

5-16 year olds; 58%

5-16 year olds:84%;

Table 2. Permanent teeth-children and adolescents; prospective studies (good level of evidence) (continued)

Researcher

Baseline Scores (Mean + SD)

True High Risk Criteria Used

Method of modeling

Country

Sampling method

Training of examiners reported

Reliability of examiners

Blinding of examiners

Blinding of patients

Subject Attrition

Authors conclusion

[Disney et al., 1992b]
North Carolina Study
"High Risk Prediction Model"

Aiken: DMFS:0.3 (grade 1), 3.0 (grade 5)

dmfs: 9.3 (grade 1), 4.4 (grade 5)

Portland;
DMFS:0.2 (grade 1), 1.7 (grade 5)

dmfs: 2.9 (grade 1), 2.4 (grade 5)

 

High risk:25% of the total sample size.

(LRA, stepwise, 38—43 variables studied)

USA

NR

Trained

Examiner reliability; intraclass correlations above 90% for 10/12 comparisons. Reliability for noncaries data showed fair agreement among examiners.

Yes

NR

Lost approx. 20% from baseline (more than N)

Models had high specificity for children at low risk. Clinical predictors were the most important ones, while the other factors contributed little to the prediction.

[Isokangas et al., 1993]

NR

High risk: Any caries increment

Not used

Finland (Ylivieska)

All 3-16 year olds in public dental care were included

15 clinicians participate. No training reported.

NR (dentists examined different children)

Not possible for ethical reasons

NR

NR

Clinicians can predict risk using only caries and sociodemographic variables available at annual examinations

Table 3. Permanent teeth adults-prospective studies (good level of evidence)

Researcher

N (dentate)

Age (t outset

Study Design

Variables: Past Caries or Disease Experience

Variables: Microflora

Variables:
Host

Variables: Other

Outcome=
Validation criteria=
true disease

Sensitivity %

Specificity %

[Hawkins et al., 1997;van Houte, 1993]

699

50+

Prospective 3 years

No calculus removed no radiographs
Third molars excluded

Mean AL (baseline)
No of teeth (baseline)
Coronal DF

Not Used

Not Used

Educational level
Marital status
Age
Total household income
Dental visiting pattern
Born in Canada
Major life event in past 6 months
Wearing partial denture

One or more net coronal DFS increments

80.2

46.2

 

Table 3. Permanent teeth adults-prospective studies (good level of evidence) (continued)

Researcher

Baseline Scores
(Mean + SD)

True High Risk Criteria Used

Method of modeling

Country

Sampling method

Training of examiners reported

Reliability of examiners

Blinding of examiners

Blinding of patients

Subject Attrition

Authors conclusion

[Hawkins et al., 1997;van Houte, 1993]

Caries incidence 57%

Mean net increment 1.91�2.60

NR

LRA

Canada, Ontario

Random

Calibration reported

94%kappa 0.76 coefficient of reproducibility 0.97 (p<0.001)

NR

NR

206

Non-clinical factors, which showed significant effects were education and marital status, both of these factors may influence attitudes towards oral health.

The baseline no. of teeth and mean periodontal AL may measure the number of tooth surfaces at risk of decay.

Conclusions

  • The predictive validity of the models reviewed depended strongly on caries prevalence and characteristics of the population on which they were based.
  • Many models included similar categories of predictors but provided very different outcomes.
  • In many instances the use of a single predictor gave results as good as those of a combination of predictors.
  • Previous caries experience was a significant predictor in most models tested for primary, permanent, and root caries.
  • The desired combination of sensitivity and specificity (more than 160 percent) was only achieved in a few cases.
  • None of the studies rated as "good" reached the desirable combined level of sensitivity + specificity.
  • None of the controlled longitudinal studies conducted to predict root caries were rated as "good."
  • Most of the research in this area has been done in children. There is, therefore, a need to develop better evidence to support caries risk assessment strategies in adults.

Future Research

Clearly, there is a need for further research to identify and validate caries risk assessment strategies that can be applied in dental practice. More importantly, studies are required to establish whether identification of high-risk individuals can lead to more effective long-term patient management that arrests or reverses the progression of carious lesions.

Another recommendation follows from the consistent finding that past caries experience is a strong predictor of future disease. Most studies have used the DMFS (decayed, missing, filled surfaces) index to determine past caries experience. This approach does not necessarily separate out the D component from the F component. Furthermore, this approach does not establish whether decayed lesions are active (progressing) or inactive (arrested). The presence of caries activity should be a much stronger predictor of future carious lesions (frank cavitations) than the DMFS index. The development of technology to detect early caries lesions and to directly assess caries lesion status may prove to be the best way to identify patients who need aggressive preventive intervention.

References

Abernathy JR, Graves RC, Bohannan HM, Stamm JW, Greenberg BG, Disney JA. Development and application of a prediction model for dental caries. Comm Dent Oral Epidemiol 1987:15:24�8.

Beck JD, Weintraub JA, Disney JA, Graves RC, Stamm JW, Kaste LM, et al., University of North Carolina Caries Risk Assessment Study: comparisons of high risk prediction, any risk prediction, and any risk etiologic models. Comm Dent Oral Epidemiol 1992;20:313�21.

Brambilla E, Gagliani M, Felloni A, Garc�a-Godoy F, Strohmenger L. Caries-preventive effect of topical amine fluoride in children with high and low salivary levels of mutans streptococci. Caries Res 1999;33:423�7.

Demers M, Brodeur JM, Mouton C, Simard PL, Trahan L, Veilleux G. A multivariate model to predict caries increment in Montreal children aged 5 years. Comm Dental Health 1992;9:273�81.

Disney JA, Graves RS, Stamm JW, Bohannan HM, Abernathy JR, Zack DD. The University of North Carolina Caries Risk Assessment study: further developments in caries risk prediction. Comm Dent Oral Epidemiol 1992;20:64�75.

Hausen H, Karkkainen S, Seppa L. Application of the high-risk strategy to control dental caries. Comm Dent Oral Epidemiol 2000;28:26�34.

Hawkins RJ, Jutai DK, Brothwell DJ, Locker D: Three-year coronal caries incidence in older Canadian adults. Caries Res 1997;31:405�10.

Holst A, Martensson I, Laurin M. Identification of caries risk children and prevention of caries in pre-school children. Swed Dent J 1997;21:185�91.

Isokangas P, Alanen P, Tiekso J. The clinician�s ability to identify caries risk subjects without saliva tests�a pilot study. Comm Dent Oral Epidemiol 1993;21:8�10.

Kingman A, Little W, Gomez I, Heifetz SB, Driscoll WS, Sheats R, et al., Salivary levels of Streptococcus mutans and lactobacilli and dental caries experiences in a US adolescent population. Comm Dent Oral Epidemiol 1988;16:98�103.

Moss ME, Zero DT. An overview of caries risk assessment and its potential utility. J Dent Educ 1995;59:932�40.

Stamm JW, Disney JA, Graves RC, Bohannan H, Abernathy JR. The University of North Carolina Caries Risk Assessment Study. I: Rationale and content. J Public Health Dent 1988;48:225�32.

Abstracts Index