Caries risk assessment (CRA) is the clinical process of establishing the probability for an individual patient to develop caries lesions in the near future and thereby an essential component in the decision-making process for adequate prevention and management of dental caries [1–3]. In comprehensive clinical practice, risk factors based on general health, diet, oral hygiene, fluoride exposure and past caries experience are often subjectively and intuitively merged into one of several risk categories [4, 5], albeit the quality of evidence for this process is limited [6]. Cariogram is an algorithm based software based on nine different caries related risk factors and intended to aid clinicians in performing more objective and consistent risk assessments [7]. The performance of the program has been validated in preschool children [8], schoolchildren [9–12] and elderly [13]. To our knowledge, the accuracy of Cariogram in young adults has only been described in one previous study with a limited sample size [14]. The aim of the present study was therefore to validate baseline caries risk classifications according to Cariogram with the actual caries development over a 3-year period in a group of young adults living in Sweden. A second aim was to compare the outcome with a caries risk assessment scheme used within Public Dental clinics (PDC) within the region of Skåne as previously described [15].

The present prospective study evaluated a computer-based caries risk assessment program in an age group seldom studied. We have earlier described the risk assessment process performed according to guidelines of the public dental clinics (PDC) in the same material [17] and the simple conclusion is that the Cariogram did not improve the accuracy of the assessments in this age group. The PDC risk assessment relied basically on past caries experience and progression of proximal enamel lesion. This concept was supported by recent systematic reviews that have suggested baseline caries prevalence as the most accurate single predictor of caries risk in all age groups [6] and unveiled limited or weak evidence for existing caries risk systems [18]. In contrast, Gao and co-workers [19] supported algorithm-based models before reasoning-based programs but their findings were based on a preschool material. The fact that the specificity was higher than the sensitivity at most cut-off levels indicated that selecting out those with low caries risk might be a more fruitful strategy than finding individuals with high risk. Notable, the Cariogram model has performed better in some previous reports, and especially in schoolchildren [10, 12]. In the study of Celik and coworkers [14], 100 young adults were followed for two years but no predictive values were reported. Nevertheless, the findings of the present study did not motivate the time and costs of saliva sampling and microbial cultivation if only the predictive values were considered. Still, the Cariogram may be helpful in patient motivation and communication of the preventive message.

Before disqualifying the Cariogram however, there are three issues that need to be addressed. At first, the disease activity in the study group was relatively low. The prevalence of new cavities was 35%, ranging from 15% in very low risk category to 55% in the very high risk category. A higher disease would likely have increased the positive predictive values in a substantial way. Secondly, the vast majority of the patients was recalled by their regular team during the course of the study and provided with various forms of preventive care. Although the Cariogram risk category was not informed, the patients were continuously made aware of their risk category according to the public dental clinics guidelines [15], and, at best, this should have influenced the treatment decisions of the dental personnel. Evidently, a successful preventive care could impair the predictive abilities of any caries risk program. Any restorative treatment decision was taken by the regular dentist and no specific recommendations or guidelines were issued for this study. Therefore, we used total caries index (DFT/DFS) rather than the D-component to reflect the three year caries increment. The third issue that could have influenced the outcome was the 24% dropout rate. Obviously, the dropouts displayed a higher burden of disease and caries risk factors and it is well known that those with active caries are most likely to develop more caries [6] and furthermore, the most caries susceptible patients were underrepresented among those that consented to the baseline examination, indicating a selection bias [15, 17]. For example, if one assumes that 80% of the dropouts would have developed new cavities over the study period, the sensitivity and positive predictive values would have increased by approximately 10%. Thus, it is important to keep in mind that the figures obtained in the present study were based on this particular population, under given circumstances, and cannot readily be generalized.

In conclusion, within the limitations of the present study, the computer-based Cariogram did not perform better than a caries risk assessment scheme based on past caries experience and caries progression, over a 3-year period in young adults.