7
Clinical Article COHORT STUDY 348 CARIES PREDICTION USING MICROBIOLOGICAL TESTS PEDIATRIC DENTISTRY V 36 / NO 4 JUL / AUG 14 Despite the advances in caries control in developed countries, dental caries remains a common chronic childhood disease. 1 In recent years, an increasing trend of caries prevalence in the primary dentition has been reported in the United States, United Kingdom, Canada, Australia, and Norway. 2-6 Further- more, the caries distribution among children is polarized, with 75 percent of affected surfaces found in less than 25 percent of children. 3,7 These epidemiological data point to the need for effective caries control in children. In particular, it would be important to proactively and accurately identify caries- susceptible children for targeted prevention and intervention and optimized treatment planning. It is widely accepted that a clinically useful caries risk assess- ment (CRA) tool should be one that is sufficiently accurate, defined as having a sum of sensitivity and specificity values (“sensitivity plus specificity”) of at least 160 percent. 8,9 Unfortu- nately, there are few prospective studies reporting that their CRA system reached such a level of accuracy, 9-11 and only two sufficiently accurate systems among preschoolers have so far been reported. 12,13 e abundance of cariogenic bacteria might be useful for predicting caries. With objective results that are meaningful for patient education, microbiological tests are considered clini- cally promising as single-factor CRAs or integrated into multi- factorial CRAs. 14 A significant association between caries and the abundance of mutans streptococci (MS; especially strepto- coccus mutans and streptococcus sobrinus ) in saliva or plaque has been demonstrated, with a pooled relative risk ( RR) of 2.11 (95% confidence interval [CI]=1.47-3.02) reported in a meta-analysis. 15 On the other hand, the association between lactobacilli (LB) and caries remains equivocal, with studies showing a positive correlation, 16 negative correlation, 17 and absence of correlation. 18,19 e sensitivity and specificity of salivary MS levels in predicting caries among preschool children ranged widely, from 33 percent to 91 percent and from 27 percent to 90 percent, respectively; however, the combined “sensi- tivity plus specificity” was no more than 130 percent. 15 Simi- larly, the reported sensitivity and specificity of the LB test ranged widely, from nine percent to 79 percent and from 50 percent to 97 percent, respectively, with a maximum “sensitivity plus specificity” of 148 percent. 20 e wide variation in reported sensitivity and specificity and low accuracy may be due to the small sample size, limited number of prospective studies, and lack of control for potential confounders. 15,21 Therefore, the clinical usefulness of these bacterial tests as a tool for predicting caries remains uncertain. e purpose of this prospective study was to systemically reappraise the potential clinical usefulness of two microbio- logical indicators (mutans streptococci and lactobacilli levels), singly and in combination, for predicting early childhood caries (ECC) and their contributions in multifactorial modeling. Methods Participants. This study was conducted among preschool children in Singapore. Ethical approval was obtained from the Institutional Review Board of the National University of Singapore (NUS-IRB 04-155), Kent Ridge, Singapore. The public water supply in Singapore is fluoridated at a level of 0.6 ppm F. e sampling frame was the People’s Action Party Community Foundation (PCF) education system, which pro- vides education to the overwhelming majority (~80 percent) of preschoolers in Singapore. A stratified random sample was obtained, proportionate to the population size of the five 1 Dr. Gao is a research assistant professor, Dental Public Health, Faculty of Dentistry, The University of Hong Kong, Hong Kong. Drs. 2 Hsu and 3 Loh are associate professors, Oral Sciences, Faculty of Dentistry; and 4 Dr. Hwarng is an associate professor, Department of Decision Sciences, School of Business; and 5 Dr. Koh is a chair professor, PAPRSB Institute of Health Sciences, Universiti Brunei Darussalam, Bandar Seri Begawan, Brunei Darussalam, and a professor, Saw Swee Hock School of Public Health, all at the National University of Singapore, Singapore. Correspond with Dr. Hsu at [email protected] Role of Microbiological Factors in Predicting Early Childhood Caries Xiaoli Gao, BDS, MSc, PhD 1 Chin-Ying Stephen Hsu, DDS, MS, PhD 2 Teresa Loh, BDS, DPHD 3 Brian Hwarng, BS, MS, PhD 4 David Koh, MBBS, MSc, PhD 5 Abstract: Purpose: Microbiological methods that accurately identify caries-susceptible children may enhance caries control and assist treat- ment planning. This study’s purpose was to evaluate the usefulness of two microbiological indicators (mutans streptococci [MS] and lactobacilli [LB] levels), singly and in combination, for predicting early childhood caries (ECC) and their contributions in multifactorial modeling. Methods: A total of 1,782 randomly selected three- to five-year-olds were recruited and underwent oral examination and microbiological tests using commercially available diagnostic kits. A questionnaire was completed by their parents. After 12 months, the caries increment in 1,576 (~88 percent) children was assessed. Results: Caries risk increased with the MS and LB levels but plateaued above the LB level of 10 5 CFU/ml saliva. MS, LB, and combined MS+LB models predicted one-year caries increment (Δdmft>0) with a sensitivity/specificity of 79 percent/67 percent, 51 percent/89 percent, and 66 percent/85 percent, respectively. Sensitivity/specificity reached 80 percent/80 percent when baseline caries experience (“past caries”) was added to the MS+LB model and up to 85 percent/80 percent and 81 percent/85 percent when psychosociobehavioral factors and oral hygiene status were added to the MS+LB model, without and with “past caries,” respectively. Conclusions: The combined “mutans streptococci+lactobacilli+past caries” model is useful for identifying at-risk children. Incorporating MS and LB into a biopsychosociobe- havioral model slightly improved the prediction, even without “past caries”. (Pediatr Dent 2014;36:348-54) Received May 21, 2013 | Last Revision July 26, 2013 | Accepted July 26, 2013 KEYWORDS: DENTAL CARIES, MUTANS STREPTOCOCCI, LACTOBACILLUS

Role of microbiological factors in predicting early childhood caries

Embed Size (px)

Citation preview

Clinical Article COHORT STUDY

348 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

Despite the advances in caries control in developed countries dental caries remains a common chronic childhood disease1 In recent years an increasing trend of caries prevalence in the primary dentition has been reported in the United States United Kingdom Canada Australia and Norway2-6 Further- more the caries distribution among children is polarized with 75 percent of affected surfaces found in less than 25 percent of children37 These epidemiological data point to the need for effective caries control in children In particular it would be important to proactively and accurately identify caries- susceptible children for targeted prevention and intervention and optimized treatment planning

It is widely accepted that a clinically useful caries risk assess-ment (CRA) tool should be one that is sufficiently accurate defined as having a sum of sensitivity and specificity values (ldquosensitivity plus specificityrdquo) of at least 160 percent89 Unfortu- nately there are few prospective studies reporting that their CRA system reached such a level of accuracy9-11 and only two sufficiently accurate systems among preschoolers have so far been reported1213

The abundance of cariogenic bacteria might be useful for predicting caries With objective results that are meaningful for patient education microbiological tests are considered clini-cally promising as single-factor CRAs or integrated into multi- factorial CRAs14 A significant association between caries and the abundance of mutans streptococci (MS especially strepto- coccus mutans and streptococcus sobrinus) in saliva or plaque

has been demonstrated with a pooled relative risk (RR) of 211 (95 confidence interval [CI]=147-302) reported in a meta-analysis15 On the other hand the association between lactobacilli (LB) and caries remains equivocal with studies showing a positive correlation16 negative correlation17 and absence of correlation1819 The sensitivity and specificity of salivary MS levels in predicting caries among preschool children ranged widely from 33 percent to 91 percent and from 27 percent to 90 percent respectively however the combined ldquosensi-tivity plus specificityrdquo was no more than 130 percent15 Simi- larly the reported sensitivity and specificity of the LB test ranged widely from nine percent to 79 percent and from 50 percent to 97 percent respectively with a maximum ldquosensitivity plus specificityrdquo of 148 percent20 The wide variation in reported sensitivity and specificity and low accuracy may be due to the small sample size limited number of prospective studies and lack of control for potential confounders1521 Therefore the clinical usefulness of these bacterial tests as a tool for predicting caries remains uncertain

The purpose of this prospective study was to systemically reappraise the potential clinical usefulness of two microbio- logical indicators (mutans streptococci and lactobacilli levels) singly and in combination for predicting early childhood caries (ECC) and their contributions in multifactorial modeling

MethodsParticipants This study was conducted among preschool children in Singapore Ethical approval was obtained from the Institutional Review Board of the National University of Singapore (NUS-IRB 04-155) Kent Ridge Singapore The public water supply in Singapore is fluoridated at a level of 06 ppm F The sampling frame was the Peoplersquos Action Party Community Foundation (PCF) education system which pro- vides education to the overwhelming majority (~80 percent) of preschoolers in Singapore A stratified random sample was obtained proportionate to the population size of the five

1Dr Gao is a research assistant professor Dental Public Health Faculty of Dentistry The University of Hong Kong Hong Kong Drs 2Hsu and 3Loh are associate professors Oral Sciences Faculty of Dentistry and 4Dr Hwarng is an associate professor Department of Decision Sciences School of Business and 5Dr Koh is a chair professor PAPRSB Institute of Health Sciences Universiti Brunei Darussalam Bandar Seri Begawan Brunei Darussalam and a professor Saw Swee Hock School of Public Health all at the National University of Singapore SingaporeCorrespond with Dr Hsu at denhsusnusedusg

Role of Microbiological Factors in Predicting Early Childhood CariesXiaoli Gao BDS MSc PhD1 bull Chin-Ying Stephen Hsu DDS MS PhD2 bull Teresa Loh BDS DPHD3 bull Brian Hwarng BS MS PhD4 bull David Koh MBBS MSc PhD5

Abstract Purpose Microbiological methods that accurately identify caries-susceptible children may enhance caries control and assist treat- ment planning This studyrsquos purpose was to evaluate the usefulness of two microbiological indicators (mutans streptococci [MS] and lactobacilli [LB] levels) singly and in combination for predicting early childhood caries (ECC) and their contributions in multifactorial modeling Methods A total of 1782 randomly selected three- to five-year-olds were recruited and underwent oral examination and microbiological tests using commercially available diagnostic kits A questionnaire was completed by their parents After 12 months the caries increment in 1576 (~88 percent) children was assessed Results Caries risk increased with the MS and LB levels but plateaued above the LB level of 105 CFUml saliva MS LB and combined MS+LB models predicted one-year caries increment (Δdmftgt0) with a sensitivityspecificity of 79 percent67 percent 51 percent89 percent and 66 percent85 percent respectively Sensitivityspecificity reached 80 percent80 percent when baseline caries experience (ldquopast cariesrdquo) was added to the MS+LB model and up to 85 percent80 percent and 81 percent85 percent when psychosociobehavioral factors and oral hygiene status were added to the MS+LB model without and with ldquopast cariesrdquo respectively Conclusions The combined ldquomutans streptococci+lactobacilli+past cariesrdquo model is useful for identifying at-risk children Incorporating MS and LB into a biopsychosociobe- havioral model slightly improved the prediction even without ldquopast cariesrdquo (Pediatr Dent 201436348-54) Received May 21 2013 | Last Revision July 26 2013 | Accepted July 26 2013

KEYWORDS DENTAL CARIES MUTANS STREPTOCOCCI LACTOBACILLUS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 349

administration districts In each district kindergartens were randomly selected from all registered PCF kindergartens in that district A total of 13 kindergartens were selected four in the central two in the southeast three in the southwest two in the northeast and two in the northwest districts One kinder- garten in the central district was not able to participate in this study due to a prescheduled health program A substitute was selected by proceeding to the next random number

All children enrolled in nursery and kindergarten-1 classes (three- to five-year-olds) were approached With parental informed consent 1782 children participated in this study The response rate was 86 percent Details of the sampling methods and subject recruitment have been described in a previous paper7

Questionnaire A self-administered questionnaire was com- pleted by parents Information was collected on their familyrsquos socioeconomic status (parentsrsquo education attainment and type of housing) childrsquos demographic background (age gender and ethnicity) oral health behaviors (feeding history diet habits oral hygiene measures fluoride exposures and dental attend- ance) and parental knowledge attitude and self-efficacy in protecting childrenrsquos teeth

Microbiological tests Before tests all children refrained from food drinks and tooth-brushing for at least one hour Each child was instructed to chew a piece of paraffin for one minute for dislodging bacteria from tooth surfaces to saliva A Dentocult SM Strip (Orion Diagnostica Espoo Finland) was pressed on hisher tongue 10 times and then removed between slightly closed lips After a further four minutes of chewing saliva was collected and spread on the agar surfaces of the Dentocult LB test slide (Orion Diagnostica) The MS and LB samples were incubated at 37 degrees Celsius for 48 and 96 hours respectively Based on the manufacturerrsquos instruction and standard charts the counts of colony forming units per milliliter (CFUml) saliva were categorized into four levels by two cali- brated examiners who were blinded to the subjectrsquos caries status Scores 0 to 3 denote an MS level of ldquo less than 104rdquo ldquo104-105rdquo ldquo105-106rdquo and ldquo greater than 106rdquo respectively and an LB level of less than 103rdquo ldquoequal to or greater than 104rdquo ldquoequal to or greater than 105rdquo and ldquogreater than 106rdquo respectively Samples from 86 children were reassessed by both examiners to evaluate the intra- and interexaminer reliability

Oral examination All children were examined by a single trained examiner at baseline and after 12 months Oral hy- giene status was evaluated using the Silness-Loumle plaque index22 Caries was registered at the cavitation level according to criteria recommended by the World Health Organization (WHO)23 Caries was considered as present when there was a cavity de- tectable softened floor or wall or undermined enamel or a surface with a temporary filling A tooth was scored as missing only if the child was at an age when normal exfoliation would not be a sufficient explanation for the absence of the tooth and other reasons for missing teeth were excluded A surface was recorded as filled when one or more permanent restorations were present If a surface had both a carious lesion and a filling it was counted as a decayed surface Owing to the limitation of the field setting initial noncavitated carious lesions were not recorded The tooth status was assessed by visual inspec- tion aided by tactile inspection if necessary No radiographs were taken Duplicate examinations were performed on 10 percent of the subjects to assess intraexaminer reliability At least 10 other children were examined between duplicate ex- aminations for each selected child

After 12 months 1576 (~88 percent) children attended a follow-up examination Information was also solicited on any dental treatment received by the participants between the base- line and follow-up examinations The examiner was blinded to the childrenrsquos microbiological test results at both the baseline and follow-up visits

Data analysis Data were analyzed using SPSS 200 soft- ware (SPSS Inc Chicago Ill USA) The intra- and interex-aminer reliability was calculated using Cohenrsquos kappa statistics The capacity of microbiological factors in predicting ECC was benchmarked against ldquopast caries experiencerdquo (baseline dmft greater than 0) which is regarded as the strongest caries indi- cator9 Through multiple logistic regressions the associations between predictive factors (MS level LB level and ldquopast cariesrdquo) and ldquoone-year caries incrementrdquo (∆dmftgt0) were evaluated after controlling for the demographic socioeconomic and behavioral variables The significance level was set at 005

On the basis of data from 50 percent of subjects randomly selected by the SPSS program models for predicting the risk of ldquoone-year caries increment (∆dmftgt0)rdquo were constructed by entering predictive factors (MS level LB level and ldquopast cariesrdquo) and their combinations Multifactorial models involving psychosociobehavioral factors oral hygiene status and one or both microbiological factors were also constructed with and without involving ldquopast cariesrdquo Data from the remaining 50 percent of subjects were used to validate and compare these models For each single factor (MS level LB level or ldquopast cariesrdquo) there were no more than three possible cutoff points each of which were explored for caries prediction For models involving two or more factors since the risk prediction is in a continuous scale receiver operation characteristics (ROC) curves were plotted for each possible cutoff point namely the predicted ldquopossibility of disease (∆dmftgt0)rdquo

Parameters used for model validation and comparison in- cluded areas under ROC curves (AUC) sensitivity (Se) speci- ficity (Sp) and accuracy Comparisons were also made with a nonmicrobiological clinical screening model reported in our previous study13 In that model only multiple psychosocio- behavioral factors (derived from the parental questionnaire) and clinical observations (oral hygiene and past caries) were used to predict risk of ECC with no microbiological tests involved

Further analysis was also performed in a subgroup ie among children who were caries free at baseline (baseline dmft=0) to evaluate (1) the association of MS and LB levels with caries increment in these children and (2) the capability of these two bacterial tests and their combination in predicting children who turned from caries free to caries affected within 12 months

Results The study sample initially comprised 1782 children (889 boys and 893 girls) including 194 888 and 697 children who were three four and five years old respectively The sample was fairly representative of the national population in terms of geographic and socioeconomic profiles7 The intra- and inter- examiner reliability for the microbiological tests was high in the range of 09 to 095 and 083 to 089 respectively For caries examination the intraexaminer reliability was 096 No significant difference was found between children lost to follow-up and those followed up in terms of their demographic and socioeconomic profiles and baseline caries status (all Pgt05) Between the baseline and follow-up examination 62 (~four percent) children had visited a dentist The treatments received

350 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

were exclusively restoration or extraction No preventive re- gimens such as fluoride applications sealants or antimicrobial therapies were recorded

At baseline 718 (40 percent) children were affected by caries The mean (plusmnSD) dmft were 154 (plusmn275) The majority (90 percent) of the affected teeth were untreated decayed teeth In 12 months 688 (44 percent) children developed new carious lesions (dmft increment Δdmftgt0 Table 1) including 206 children who were cavity-free at baseline and 482 children who had cavities at baseline and developed a new cavity on at least one originally intact tooth High MS levels (Score 2 or 3 ge105 CFUml saliva)24 and high LB levels (score 1-3 ge104 CFUml saliva)16 were found in 53 percent and 28 percent of children respectively

Associations between each predictive factor and new caries (dmft increment Δdmftgt0) were expressed as relative risk (RR) after controlling for sociodemographic and behavioral factors (Table 1) Compared with children having an MS score of 0 those with scores of 1 2 and 3 were significantly more likely to have new caries as reflected by an RR (95 CI) of 196

(126-283) 341 (259-419) and 464 (389-523) respectively Compared with children whose LB score was 0 those scored as 1 2 and 3 had an RR (95 CI) of 193 (144-237) 272 (221-302) and 270 (223-299) respectively Compared with children who were cavity-free at baseline those found with any cavity at the start of the study (baseline dmftgt0) were 162 (95 CI=125-190) times as likely to develop caries in 12 months Among children who were caries free at baseline the mean caries increment percent with new caries and RR for each stratum was lower vs statistics from the whole sample

Multifactorial CRA models involving MS LB and both are described in Table 2 Other variables included in one or more of these models were age ethnicity parental educa-tion breast-feeding history intake of sweets use of fluoride- containing toothpaste systemic health past caries and oral hygiene status The performances of these multifactorial models single factors (MS LB and ldquopast cariesrdquo) and their combinations in assessing childrenrsquos caries risk are listed and compared in Table 3 The optimal cutoff point of the MS level for predicting caries was ldquoCFUml saliva ge105 (score 2)rdquo

Table 1 CARIES INCREMENT IN CHILDREN WITH DIFFERENT PAST CARIES EXPERIENCE AND DIFFERENT BACTERIAL LEVELS

Factors n Mean (SD)caries increment

(Δdmft)

withnew caries(Δdmftgt0)

RR (95 CI)dagger P-value

MS level (all children) Dentocult score 0 (lt104 CFUml saliva) 533 34 030 (084)I 16 1 (reference) lt001

Dentocult score 1 (104 -105 CFUml saliva)

198 13 058 (131)I 28 196 (126-283)

Dentocult score 2 (105 -106 CFUml saliva)

382 25 111 (150)II 53 341 (259-419)

Dentocult score 3 (gt106 CFUml saliva) 445 29 165 (153)III 75 464 (389-523)

MS level (children with baseline dmft=0)

Dentocult score 0 (lt104 CFUml saliva) 464 50 022 (073) 110 1 (reference)

Dentocult score 1 (104 -105 CFUml saliva)

143 15 038 (121) 182 177 (129-206)

Dentocult score 2 (105 -106 CFUml saliva)

195 21 056 (113) 282 200 (172-216)

Dentocult score 3 (gt106 CFUml saliva) 133 14 105 (138) 534 219 (204-225)

LB level (all children) Dentocult score 0 (103 CFUml saliva) 1029 72 061 (123)a 30 1 (reference) lt001

Dentocult score 1 (104 CFUml saliva) 148 10 130 (132)b 69 193 (144-237)

Dentocult score 2 (105 CFUml saliva) 99 7 194 (177)c 81 272 (221-302)

Dentocult score 3 (106 CFUml saliva) 160 11 201 (141)c 88 270 (223-299)

LB level (children with baseline dmft=0)

Dentocult score 0 (103 CFUml saliva) 779 92 038 (102) 19 1 (reference)

Dentocult score 1 (104 CFUml saliva) 31 4 058 (085) 39 166 (070-215)

Dentocult score 2 (105 CFUml saliva) 26 3 162 (158) 65 202 (056-228)

Dentocult score 3 (106 CFUml saliva) 14 2 186 (117) 79 216 (156-227)

Past cariesDagger Cavity-free (baseline dmft=0) 940 60 044 (105)i 22 1 (reference) lt001

With cavity (baseline dmftgt0) 636 40 166 (157)ii 76 162 (125-190)

Total 1576 100 093 (144) 44

Ranking (I-III i-ii and a-c) there was significant difference in mean caries increment (∆dmft) between strata with different ranks Tukeyrsquos post-hoc tests or independent t tests (as appropriate) were used to compare means while Kruskal-Wallis tests or Mann-Whitney tests (as appropriate) were employed when the normal distribution or homo- geneity of variance was violateddagger Odds Ratios (OR) and their 95 confidence intervals (CI) were derived from multiple logistic regressions after adjusting for demographic socioeconomic and behavioral factors For a better reflection of the risk in this prospective cohort study ORs were converted to relative risk (RR) thorough the following equation RR=OR[(1-P0) + (P0 X OR)] (P0 stands for the disease incidence)Dagger Both microbiological factors were benchmarked against ldquopast cariesrdquo which is regarded as the strongest caries indicator9

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 351

reaching a sensitivityspecificity of 79 percent67 percent For the LB level the optimal cutoff point was score 1 (ge104 CFUml saliva) with a sensitivityspecificity of 51 percent89 percent in predicting caries When MS was combined with LB the sensitivityspecificity (66 percent85 percent) was similar to that of ldquopast cariesrdquo (70 percent83 percent) Adding LB or MS to ldquopast cariesrdquo minimally increased its sen- sitivity or specificity When both microbiological factors (MS and LB) were combined with ldquopast cariesrdquo the sensitivity and specificity totaled 160 percent with each being 80 percent

The nonmicrobiological clinical screening model that included only psychobehavioral questionnaire in- formation and clinical examination results (oral hygiene status and ldquopast cariesrdquo) yielded an AUC close to that of the combined ldquoMS+LB+past cariesrdquo model (0845 vs 0855) However its specificity was slightly lower (73 percent vs 80 percent) Incorporating only the LB test into that clinical screening model improved the spe- cificity considerably (73-85 percent) with a slight decrease in sensitivity (82-78 percent) When both MS and LB were added to the multifactorial model signifi- cantly higher overall performances (AUC 0845-0897) were observed with a sensitivityspecificity of 81 percent 85 percent Subtraction of ldquopast cariesrdquo from the mutifac- torial models revealed that this factor contributed mini- mally in the multifactorial models involving MS (alone or with LB) Without ldquopast cariesrdquo the multifactorial model involving MS and LB reached a sensitivity specificity of 85 percent80 percent The ROC curves of combinations of factors and multifactorial models are presented in Figure 1 paralleling the aforementioned findings

In the subgroup of children who were caries free at baseline the low specificity of MS level and low sensiti- vity of LB level in predicting caries was further exacer- bated When both tests were combined the SeSp was 68 percent72 percent with a summation of 139 percent Compared with the corresponding statistics in the whole sample (SeSp=66 percent85 percent) a drop in specificity was observed indicating more false positive cases in this sub- group of children

DiscussionMethodological considerations This popula-tion-based prospective study in a relatively large sample aimed to generate new evidence on the usefulness of microbiological parameters in predicting ECC as single factors in combina- tion and in multifactorial modeling A limita- tion of this study was that initial noncavitated carious lesions were not registered The findings of this study are therefore better understood as appraisal of caries prediction methods at the cavitation level Although 62 (four percent) children had visited a dentist during the study period the types of dental care they had re- ceived were exclusively curative (restoration and extraction) without any preventive measures Therefore the confounding impacts of these dental visits may be negligible Commercially

available diagnostic kits were used to evaluate the level of cario- genic bacteria in this study Although their accuracy may be less ideal compared with advanced laboratory technologies

Odds ratios and their 95 confidence intervals (CI) were derived logistic regressions The out- come variable was ldquoany caries incrementrdquo in a year (∆dmft gt0 or =0)

Table 2 MULTIFACTORIAL CARIES PREDICTION MODELS INVOLVING MUTANS STREPTOCOCCI (MS) ANDOR LACTOBACILLI (LB)

Odds ratio (95 CI)

Multifactorial model with MS

Multifactorial model with LB

Multifactorial model with MS

and LB

Questionnaire informationAge (mos) 1056 (1029-1084) 1041 (1015-1067) 1051 (1023-1078)

Malay race 1764 (1096-2841)

Fatherrsquos education

0668 (0552-0809) 0630 (0533-0744) 0636 (0527-0738)

Months of breastfeeding

1045 (1017-1074) 1040 (1016-1065) 1059 (1030-1087)

Frequency of sweet

1393 (1112-1746)

Fluoridated toothpaste

0580 (0346-0973) 0554 (0340-0903) 0573 (0343-0953)

No health problems

2436 (1428-4156) 2352 (1451-3815) 2153 (1263-3618)

Clinical dataPast (baseline) caries

4279 (2899-6314) 4820 (3285-7073) 2983 (1698-5497)

Plaque index 5205 (3227-8394) 5212 (3350-8109) 5197(3220-8385)

Biological testsMS 2161 (1839-2539) 2104 (1782-2486)

LB 1932 (1549-2409) 1859 (1161-2998)

Figure 1 Receiver operation characteristics (ROC) curves for models involving two or more factors

352 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

The performance of our previously reported clinical screening model and past caries which is regarded as the strongest predictor for future caries is pre- sented in this table for comparison The clinical screening model requires information on sociodemographic background oral health behaviors and clinical conditions (oral hygiene and past caries)13

dagger Receiver operation characteristics (ROC) curves were only plotted when there are multiple possible cutoff points Area under ROC curve (AUC) were there- fore only available for these modelsDagger Cutoff point with highest ldquosensitivity (Se) + specificity (Sp)rdquo in indicating caries Data under these optimal cutoff points were bolded for easy comparisonsRanking (I-II) There was a significant difference in AUC between factorscombinationsmodels with different ranks

(immunoassay and agar-based quantification) they are validated and easy to use require little apparatus and are currently the only common and practical option in the clinical setting

Dental caries appeared to be more prevalent in Singaporean children than among children in Western countries This is in line with the ECC pattern in Asian populations1 The skewed distribution of caries lesions in this sample however justifies the

need for identifying high-risk children for intensive prevention and intervention

Single microbiological factors and their combinations As a single factor both MS and LB achieved an accuracy of 72 percent in predicting ECCmdashlower than that of ldquopast cariesrdquo (77 percent) Although there was a positive gradient between MS level and ECC caries risk approached a plateau when the

Table 3 PREDICTIVE VALUES USING SINGLE FACTORS (MUTANS STREPTOCOCCI [MS] LACTOBACILLI [LB] AND ldquoPAST CARIESrdquo) THEIR COMBINATIONS AND MULTIFACTORIAL MODELS

Screening criteria for children with any risk (Δdmftgt0)

Performance measures

AUCdagger

Se Sp Se+Sp Accuracy

Single factorsPast caries Baseline dmft gt0 70 83 153 77 NA

MS level (all children) Dentocult score ge1 (ge104 CFUml saliva) 88 51 139 67 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 79 67 146 72

Dentocult score ge3 (gt106 CFUml saliva) 50 88 137 71

MS level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva) 77 43 120 58 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 64 27 91 43

Dentocult score ge3 (gt106 CFUml saliva) 35 7 42 19

LB level (all children) Dentocult score ge1 (ge104 CFUml saliva)Dagger 51 89 140 72 NA

Dentocult score ge2 (ge105 CFUml saliva) 35 95 130 69

Dentocult score ge3 (gt106 CFUml saliva) 22 98 120 65

LB level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva)Dagger 21 95 117 63 NA

Dentocult score ge2 (ge105 CFUml saliva) 15 98 113 62

Dentocult score ge3 (gt106 CFUml saliva) 6 99 105 58

CombinationsMS + LB (all children) Predicted ldquopossibility of diseaserdquo ge0524 66 85 151 77 0822II

MS + LB (children with baseline dmft=0)

Predicted ldquopossibility of diseaserdquo ge 0239 68 72 139 70 0788

MS + past caries Predicted ldquopossibility of diseaserdquo ge0380 81 77 158 79 0841II

LB + past caries Predicted ldquopossibility of diseaserdquo ge0406 75 81 156 78 0805II

MS + LB + past caries Predicted ldquopossibility of diseaserdquo ge0435 80 80 160 80 0855III

Multifactorial modelClinical screening model Predicted ldquopossibility of diseaserdquo ge0328 82 73 155 77 0845II

Without past caries Predicted ldquopossibility of diseaserdquo ge0416 75 76 151 75 0803II

Multifactorial model (MS) Predicted ldquopossibility of diseaserdquo ge0362 85 78 164 81 0888III

Without past caries Predicted ldquopossibility of diseaserdquo ge0385 85 77 162 81 0875III

Multifactorial model (LB) Predicted ldquopossibility of diseaserdquo ge0434 78 85 163 82 0862III

Without past caries Predicted ldquopossibility of diseaserdquo ge0324 83 72 155 77 0850III

Multifactorial model (MS + LB)

Predicted ldquopossibility of diseaserdquo ge0352 81 85 166 83 0897I

Without past caries Predicted ldquopossibility of diseaserdquo ge0280 85 80 165 82 0889III

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 353

LB level reached ldquoScore 2rdquo It is possible that an increased LB count includes an increase in the level of probiotic LB species balancing out the effect of cariogenic species Such an assump- tion is supported by studies showing that many Lactobacillus species in saliva significantly inhibit the growth of MS2526 Nevertheless since our study did not include a quantification of different LB species this assumption is yet to be tested

It has been suggested that LB may be a better caries pre- dictor compared with MS count27 Our study showed no signifi- cant difference in the ldquosensitivity plus specificityrdquo value between MS (146 percent) and LB (140 percent) However MS achieved a significantly higher sensitivity (79 percent vs 51 percent) whereas LB showed a higher specificity (89 percent vs 67 percent) This substantiated previously published data collected from preschoolers2829 Adding LB to MS without a third factor involved appeared to be of little value if one considers the similar Se+Sp (146 percent vs 151 percent) of MS model and MS+LB model and the additional cost spent on the LB test Among children who were caries free at baseline the combina- tion of MS and LB could only generate a SeSp of 68 percent 72 percent showing limited potential of these two bacterial tests in identifying caries-free children for early prevention

An ldquoMS+past cariesrdquo model has been chosen by some re- searchers as a simple screening tool for selecting children with high caries risk3031 However our findings suggested that adding MS to ldquopast cariesrdquo did not significantly improve accuracy (from 77-79 percent) or AUC Only after the addition of both MS and LB levels did the prediction accuracy of ldquopast cariesrdquo im- prove with a sensitivityspecificity of 80 percent80 percent and a significantly higher AUC (0855 versus 0783 Plt05)

Role of microbiological factors in multifactorial models It is encouraging that the nonmicrobiological clinical screening model (sensitivityspecificity of 82 percent73 percent) achieved a similar performance to the model combining microbiological tests and ldquopast cariesrdquo (80 percent80 percent) This highlights the importance of proper history taking and oral examination in identifying at-risk preschoolers Incorporating either MS or LB tests into the multifactorial model elevated the summation of sensitivity and specificity to above 160 percent with no sig- nificant improvement in AUC Only when both MS and LB tests were incorporated was the AUC significantly increased (0889 versus 0845 Plt05) with balanced sensitivity (85 percent) and specificity (80 percent) The incorporation of MS and LB tests into the multifactorial model mainly improved the Sp (73-85 percent) with minimal changes in Se (82-81 per- cent) This improvement in specificity is of particular impor- tance for countries with relatively low caries rates where false positive diagnosis may thus incur considerable overtreatment and waste of resources

A recent study showed a sharp decrease in specificity (from 60 percent to 47 percent) when MS was excluded from a multi- factorial CRA program called Cariogram32 However such a drastic decrease did not appear in our data for the multifac- torial model (sensitivityspecificity 85 percent78 percent with MS versus 82 percent73 percent without MS) seemingly con- firming the critical contribution of sociodemographic factors in our models and the related weighting system established through our relatively large-scale longitudinal data13

When ldquopast cariesrdquo was excluded from the modeling multi-factorial models involving MS still predicted ECC adequately reaching ldquosensitivity plus specificityrdquo values of 165 percent and 162 percent with and without LB respectively This result im- plies that MS and ldquopast cariesrdquo may be playing a similar role in

the multifactorial model involving sociodemographic behav-ioral and clinical factors Although ldquopast cariesrdquo is regarded as the best indicator for future caries this variable actually reflects the ldquoconsequencerdquo and not the ldquocauserdquo of caries Replacing ldquopast cariesrdquo with a modifiable MS factor may yield a multifacto-rial model that could be effectively utilized for patient or parent education monitoring disease activity and motivation building during follow-up visits Although the current cost to quantify MS and LB is not inexpensive the advancing technologies might substantially reduce the overall cost when the applica- tion of these tests becomes widely accepted and the commercial value is demonstrated as has happened with other health care products and services

This studyrsquos findings were generated from a child popula- tion fully covered by a fluoridated public water supply Further investigations may be needed for understanding the role of microbiological factors in predicting ECC in nonfluoridated communities

ConclusionsBased on the findings of this study the following conclusions can be drawn

1 The combined lsquomutans streptococci + lactobacilli + past cariesrsquo model with both sensitivity and specifi- city of 80 percent may serve as a useful method for selecting at-risk children for targeted intervention

2 Proper history taking and oral examination remain pertinent in identifying at risk preschoolers especially in rural or financially disadvantaged communities

3 When resources allow incorporating MS and LB into a biopsychosociobehavioral model may offer accurate prediction of ECC among preschoolers even without lsquopast cariesrsquo status

AcknowledgmentsThe authors wish to thank the participating kindergartens for their kind support Dr Robert Yee for his valuable comments on the manuscript and Dr Trevor Lane for his editorial assist- ance This study was financially supported by the Singapore Ministry of Education Academic Research Funds R222-000-021-112 and R222-000-022-112 The funders had no role in study design data collection and analysis decision to publish or preparation of the manuscript

References1 World Health Organization Oral Health CountryArea

Profile Programme (CAPP) Geneva WHO 2006 Avail- able at ldquohttpwwwmahsecapprdquo Accessed June 12 2014

2 US Department of Health and Human Services Trend in Oral Health Status United States 1988-1994 And 1999-2004 Atlanta Ga Centers for Disease Control and Prevention National Center for Health Statistics 2007

3 Pitts NB Chestnutt IG Evans D White D Chadwick B Steele JG The dentinal caries experience of children in the United Kingdom 2003 Br Dent J 200625313-20

4 Speechley M Johnston DW Some evidence from Ontario Canada of a reversal in the dental caries decline Caries Res 199630423-7

5 Armfield JM Roberts-Thomson KF Slade GD Spencer AJ Dental Health Differences between Boys and Girls The Child Dental Health Survey Australia 2000 AIHW cat no DEN 131 Canberra Australia Australian Institute

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 349

administration districts In each district kindergartens were randomly selected from all registered PCF kindergartens in that district A total of 13 kindergartens were selected four in the central two in the southeast three in the southwest two in the northeast and two in the northwest districts One kinder- garten in the central district was not able to participate in this study due to a prescheduled health program A substitute was selected by proceeding to the next random number

All children enrolled in nursery and kindergarten-1 classes (three- to five-year-olds) were approached With parental informed consent 1782 children participated in this study The response rate was 86 percent Details of the sampling methods and subject recruitment have been described in a previous paper7

Questionnaire A self-administered questionnaire was com- pleted by parents Information was collected on their familyrsquos socioeconomic status (parentsrsquo education attainment and type of housing) childrsquos demographic background (age gender and ethnicity) oral health behaviors (feeding history diet habits oral hygiene measures fluoride exposures and dental attend- ance) and parental knowledge attitude and self-efficacy in protecting childrenrsquos teeth

Microbiological tests Before tests all children refrained from food drinks and tooth-brushing for at least one hour Each child was instructed to chew a piece of paraffin for one minute for dislodging bacteria from tooth surfaces to saliva A Dentocult SM Strip (Orion Diagnostica Espoo Finland) was pressed on hisher tongue 10 times and then removed between slightly closed lips After a further four minutes of chewing saliva was collected and spread on the agar surfaces of the Dentocult LB test slide (Orion Diagnostica) The MS and LB samples were incubated at 37 degrees Celsius for 48 and 96 hours respectively Based on the manufacturerrsquos instruction and standard charts the counts of colony forming units per milliliter (CFUml) saliva were categorized into four levels by two cali- brated examiners who were blinded to the subjectrsquos caries status Scores 0 to 3 denote an MS level of ldquo less than 104rdquo ldquo104-105rdquo ldquo105-106rdquo and ldquo greater than 106rdquo respectively and an LB level of less than 103rdquo ldquoequal to or greater than 104rdquo ldquoequal to or greater than 105rdquo and ldquogreater than 106rdquo respectively Samples from 86 children were reassessed by both examiners to evaluate the intra- and interexaminer reliability

Oral examination All children were examined by a single trained examiner at baseline and after 12 months Oral hy- giene status was evaluated using the Silness-Loumle plaque index22 Caries was registered at the cavitation level according to criteria recommended by the World Health Organization (WHO)23 Caries was considered as present when there was a cavity de- tectable softened floor or wall or undermined enamel or a surface with a temporary filling A tooth was scored as missing only if the child was at an age when normal exfoliation would not be a sufficient explanation for the absence of the tooth and other reasons for missing teeth were excluded A surface was recorded as filled when one or more permanent restorations were present If a surface had both a carious lesion and a filling it was counted as a decayed surface Owing to the limitation of the field setting initial noncavitated carious lesions were not recorded The tooth status was assessed by visual inspec- tion aided by tactile inspection if necessary No radiographs were taken Duplicate examinations were performed on 10 percent of the subjects to assess intraexaminer reliability At least 10 other children were examined between duplicate ex- aminations for each selected child

After 12 months 1576 (~88 percent) children attended a follow-up examination Information was also solicited on any dental treatment received by the participants between the base- line and follow-up examinations The examiner was blinded to the childrenrsquos microbiological test results at both the baseline and follow-up visits

Data analysis Data were analyzed using SPSS 200 soft- ware (SPSS Inc Chicago Ill USA) The intra- and interex-aminer reliability was calculated using Cohenrsquos kappa statistics The capacity of microbiological factors in predicting ECC was benchmarked against ldquopast caries experiencerdquo (baseline dmft greater than 0) which is regarded as the strongest caries indi- cator9 Through multiple logistic regressions the associations between predictive factors (MS level LB level and ldquopast cariesrdquo) and ldquoone-year caries incrementrdquo (∆dmftgt0) were evaluated after controlling for the demographic socioeconomic and behavioral variables The significance level was set at 005

On the basis of data from 50 percent of subjects randomly selected by the SPSS program models for predicting the risk of ldquoone-year caries increment (∆dmftgt0)rdquo were constructed by entering predictive factors (MS level LB level and ldquopast cariesrdquo) and their combinations Multifactorial models involving psychosociobehavioral factors oral hygiene status and one or both microbiological factors were also constructed with and without involving ldquopast cariesrdquo Data from the remaining 50 percent of subjects were used to validate and compare these models For each single factor (MS level LB level or ldquopast cariesrdquo) there were no more than three possible cutoff points each of which were explored for caries prediction For models involving two or more factors since the risk prediction is in a continuous scale receiver operation characteristics (ROC) curves were plotted for each possible cutoff point namely the predicted ldquopossibility of disease (∆dmftgt0)rdquo

Parameters used for model validation and comparison in- cluded areas under ROC curves (AUC) sensitivity (Se) speci- ficity (Sp) and accuracy Comparisons were also made with a nonmicrobiological clinical screening model reported in our previous study13 In that model only multiple psychosocio- behavioral factors (derived from the parental questionnaire) and clinical observations (oral hygiene and past caries) were used to predict risk of ECC with no microbiological tests involved

Further analysis was also performed in a subgroup ie among children who were caries free at baseline (baseline dmft=0) to evaluate (1) the association of MS and LB levels with caries increment in these children and (2) the capability of these two bacterial tests and their combination in predicting children who turned from caries free to caries affected within 12 months

Results The study sample initially comprised 1782 children (889 boys and 893 girls) including 194 888 and 697 children who were three four and five years old respectively The sample was fairly representative of the national population in terms of geographic and socioeconomic profiles7 The intra- and inter- examiner reliability for the microbiological tests was high in the range of 09 to 095 and 083 to 089 respectively For caries examination the intraexaminer reliability was 096 No significant difference was found between children lost to follow-up and those followed up in terms of their demographic and socioeconomic profiles and baseline caries status (all Pgt05) Between the baseline and follow-up examination 62 (~four percent) children had visited a dentist The treatments received

350 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

were exclusively restoration or extraction No preventive re- gimens such as fluoride applications sealants or antimicrobial therapies were recorded

At baseline 718 (40 percent) children were affected by caries The mean (plusmnSD) dmft were 154 (plusmn275) The majority (90 percent) of the affected teeth were untreated decayed teeth In 12 months 688 (44 percent) children developed new carious lesions (dmft increment Δdmftgt0 Table 1) including 206 children who were cavity-free at baseline and 482 children who had cavities at baseline and developed a new cavity on at least one originally intact tooth High MS levels (Score 2 or 3 ge105 CFUml saliva)24 and high LB levels (score 1-3 ge104 CFUml saliva)16 were found in 53 percent and 28 percent of children respectively

Associations between each predictive factor and new caries (dmft increment Δdmftgt0) were expressed as relative risk (RR) after controlling for sociodemographic and behavioral factors (Table 1) Compared with children having an MS score of 0 those with scores of 1 2 and 3 were significantly more likely to have new caries as reflected by an RR (95 CI) of 196

(126-283) 341 (259-419) and 464 (389-523) respectively Compared with children whose LB score was 0 those scored as 1 2 and 3 had an RR (95 CI) of 193 (144-237) 272 (221-302) and 270 (223-299) respectively Compared with children who were cavity-free at baseline those found with any cavity at the start of the study (baseline dmftgt0) were 162 (95 CI=125-190) times as likely to develop caries in 12 months Among children who were caries free at baseline the mean caries increment percent with new caries and RR for each stratum was lower vs statistics from the whole sample

Multifactorial CRA models involving MS LB and both are described in Table 2 Other variables included in one or more of these models were age ethnicity parental educa-tion breast-feeding history intake of sweets use of fluoride- containing toothpaste systemic health past caries and oral hygiene status The performances of these multifactorial models single factors (MS LB and ldquopast cariesrdquo) and their combinations in assessing childrenrsquos caries risk are listed and compared in Table 3 The optimal cutoff point of the MS level for predicting caries was ldquoCFUml saliva ge105 (score 2)rdquo

Table 1 CARIES INCREMENT IN CHILDREN WITH DIFFERENT PAST CARIES EXPERIENCE AND DIFFERENT BACTERIAL LEVELS

Factors n Mean (SD)caries increment

(Δdmft)

withnew caries(Δdmftgt0)

RR (95 CI)dagger P-value

MS level (all children) Dentocult score 0 (lt104 CFUml saliva) 533 34 030 (084)I 16 1 (reference) lt001

Dentocult score 1 (104 -105 CFUml saliva)

198 13 058 (131)I 28 196 (126-283)

Dentocult score 2 (105 -106 CFUml saliva)

382 25 111 (150)II 53 341 (259-419)

Dentocult score 3 (gt106 CFUml saliva) 445 29 165 (153)III 75 464 (389-523)

MS level (children with baseline dmft=0)

Dentocult score 0 (lt104 CFUml saliva) 464 50 022 (073) 110 1 (reference)

Dentocult score 1 (104 -105 CFUml saliva)

143 15 038 (121) 182 177 (129-206)

Dentocult score 2 (105 -106 CFUml saliva)

195 21 056 (113) 282 200 (172-216)

Dentocult score 3 (gt106 CFUml saliva) 133 14 105 (138) 534 219 (204-225)

LB level (all children) Dentocult score 0 (103 CFUml saliva) 1029 72 061 (123)a 30 1 (reference) lt001

Dentocult score 1 (104 CFUml saliva) 148 10 130 (132)b 69 193 (144-237)

Dentocult score 2 (105 CFUml saliva) 99 7 194 (177)c 81 272 (221-302)

Dentocult score 3 (106 CFUml saliva) 160 11 201 (141)c 88 270 (223-299)

LB level (children with baseline dmft=0)

Dentocult score 0 (103 CFUml saliva) 779 92 038 (102) 19 1 (reference)

Dentocult score 1 (104 CFUml saliva) 31 4 058 (085) 39 166 (070-215)

Dentocult score 2 (105 CFUml saliva) 26 3 162 (158) 65 202 (056-228)

Dentocult score 3 (106 CFUml saliva) 14 2 186 (117) 79 216 (156-227)

Past cariesDagger Cavity-free (baseline dmft=0) 940 60 044 (105)i 22 1 (reference) lt001

With cavity (baseline dmftgt0) 636 40 166 (157)ii 76 162 (125-190)

Total 1576 100 093 (144) 44

Ranking (I-III i-ii and a-c) there was significant difference in mean caries increment (∆dmft) between strata with different ranks Tukeyrsquos post-hoc tests or independent t tests (as appropriate) were used to compare means while Kruskal-Wallis tests or Mann-Whitney tests (as appropriate) were employed when the normal distribution or homo- geneity of variance was violateddagger Odds Ratios (OR) and their 95 confidence intervals (CI) were derived from multiple logistic regressions after adjusting for demographic socioeconomic and behavioral factors For a better reflection of the risk in this prospective cohort study ORs were converted to relative risk (RR) thorough the following equation RR=OR[(1-P0) + (P0 X OR)] (P0 stands for the disease incidence)Dagger Both microbiological factors were benchmarked against ldquopast cariesrdquo which is regarded as the strongest caries indicator9

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 351

reaching a sensitivityspecificity of 79 percent67 percent For the LB level the optimal cutoff point was score 1 (ge104 CFUml saliva) with a sensitivityspecificity of 51 percent89 percent in predicting caries When MS was combined with LB the sensitivityspecificity (66 percent85 percent) was similar to that of ldquopast cariesrdquo (70 percent83 percent) Adding LB or MS to ldquopast cariesrdquo minimally increased its sen- sitivity or specificity When both microbiological factors (MS and LB) were combined with ldquopast cariesrdquo the sensitivity and specificity totaled 160 percent with each being 80 percent

The nonmicrobiological clinical screening model that included only psychobehavioral questionnaire in- formation and clinical examination results (oral hygiene status and ldquopast cariesrdquo) yielded an AUC close to that of the combined ldquoMS+LB+past cariesrdquo model (0845 vs 0855) However its specificity was slightly lower (73 percent vs 80 percent) Incorporating only the LB test into that clinical screening model improved the spe- cificity considerably (73-85 percent) with a slight decrease in sensitivity (82-78 percent) When both MS and LB were added to the multifactorial model signifi- cantly higher overall performances (AUC 0845-0897) were observed with a sensitivityspecificity of 81 percent 85 percent Subtraction of ldquopast cariesrdquo from the mutifac- torial models revealed that this factor contributed mini- mally in the multifactorial models involving MS (alone or with LB) Without ldquopast cariesrdquo the multifactorial model involving MS and LB reached a sensitivity specificity of 85 percent80 percent The ROC curves of combinations of factors and multifactorial models are presented in Figure 1 paralleling the aforementioned findings

In the subgroup of children who were caries free at baseline the low specificity of MS level and low sensiti- vity of LB level in predicting caries was further exacer- bated When both tests were combined the SeSp was 68 percent72 percent with a summation of 139 percent Compared with the corresponding statistics in the whole sample (SeSp=66 percent85 percent) a drop in specificity was observed indicating more false positive cases in this sub- group of children

DiscussionMethodological considerations This popula-tion-based prospective study in a relatively large sample aimed to generate new evidence on the usefulness of microbiological parameters in predicting ECC as single factors in combina- tion and in multifactorial modeling A limita- tion of this study was that initial noncavitated carious lesions were not registered The findings of this study are therefore better understood as appraisal of caries prediction methods at the cavitation level Although 62 (four percent) children had visited a dentist during the study period the types of dental care they had re- ceived were exclusively curative (restoration and extraction) without any preventive measures Therefore the confounding impacts of these dental visits may be negligible Commercially

available diagnostic kits were used to evaluate the level of cario- genic bacteria in this study Although their accuracy may be less ideal compared with advanced laboratory technologies

Odds ratios and their 95 confidence intervals (CI) were derived logistic regressions The out- come variable was ldquoany caries incrementrdquo in a year (∆dmft gt0 or =0)

Table 2 MULTIFACTORIAL CARIES PREDICTION MODELS INVOLVING MUTANS STREPTOCOCCI (MS) ANDOR LACTOBACILLI (LB)

Odds ratio (95 CI)

Multifactorial model with MS

Multifactorial model with LB

Multifactorial model with MS

and LB

Questionnaire informationAge (mos) 1056 (1029-1084) 1041 (1015-1067) 1051 (1023-1078)

Malay race 1764 (1096-2841)

Fatherrsquos education

0668 (0552-0809) 0630 (0533-0744) 0636 (0527-0738)

Months of breastfeeding

1045 (1017-1074) 1040 (1016-1065) 1059 (1030-1087)

Frequency of sweet

1393 (1112-1746)

Fluoridated toothpaste

0580 (0346-0973) 0554 (0340-0903) 0573 (0343-0953)

No health problems

2436 (1428-4156) 2352 (1451-3815) 2153 (1263-3618)

Clinical dataPast (baseline) caries

4279 (2899-6314) 4820 (3285-7073) 2983 (1698-5497)

Plaque index 5205 (3227-8394) 5212 (3350-8109) 5197(3220-8385)

Biological testsMS 2161 (1839-2539) 2104 (1782-2486)

LB 1932 (1549-2409) 1859 (1161-2998)

Figure 1 Receiver operation characteristics (ROC) curves for models involving two or more factors

352 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

The performance of our previously reported clinical screening model and past caries which is regarded as the strongest predictor for future caries is pre- sented in this table for comparison The clinical screening model requires information on sociodemographic background oral health behaviors and clinical conditions (oral hygiene and past caries)13

dagger Receiver operation characteristics (ROC) curves were only plotted when there are multiple possible cutoff points Area under ROC curve (AUC) were there- fore only available for these modelsDagger Cutoff point with highest ldquosensitivity (Se) + specificity (Sp)rdquo in indicating caries Data under these optimal cutoff points were bolded for easy comparisonsRanking (I-II) There was a significant difference in AUC between factorscombinationsmodels with different ranks

(immunoassay and agar-based quantification) they are validated and easy to use require little apparatus and are currently the only common and practical option in the clinical setting

Dental caries appeared to be more prevalent in Singaporean children than among children in Western countries This is in line with the ECC pattern in Asian populations1 The skewed distribution of caries lesions in this sample however justifies the

need for identifying high-risk children for intensive prevention and intervention

Single microbiological factors and their combinations As a single factor both MS and LB achieved an accuracy of 72 percent in predicting ECCmdashlower than that of ldquopast cariesrdquo (77 percent) Although there was a positive gradient between MS level and ECC caries risk approached a plateau when the

Table 3 PREDICTIVE VALUES USING SINGLE FACTORS (MUTANS STREPTOCOCCI [MS] LACTOBACILLI [LB] AND ldquoPAST CARIESrdquo) THEIR COMBINATIONS AND MULTIFACTORIAL MODELS

Screening criteria for children with any risk (Δdmftgt0)

Performance measures

AUCdagger

Se Sp Se+Sp Accuracy

Single factorsPast caries Baseline dmft gt0 70 83 153 77 NA

MS level (all children) Dentocult score ge1 (ge104 CFUml saliva) 88 51 139 67 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 79 67 146 72

Dentocult score ge3 (gt106 CFUml saliva) 50 88 137 71

MS level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva) 77 43 120 58 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 64 27 91 43

Dentocult score ge3 (gt106 CFUml saliva) 35 7 42 19

LB level (all children) Dentocult score ge1 (ge104 CFUml saliva)Dagger 51 89 140 72 NA

Dentocult score ge2 (ge105 CFUml saliva) 35 95 130 69

Dentocult score ge3 (gt106 CFUml saliva) 22 98 120 65

LB level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva)Dagger 21 95 117 63 NA

Dentocult score ge2 (ge105 CFUml saliva) 15 98 113 62

Dentocult score ge3 (gt106 CFUml saliva) 6 99 105 58

CombinationsMS + LB (all children) Predicted ldquopossibility of diseaserdquo ge0524 66 85 151 77 0822II

MS + LB (children with baseline dmft=0)

Predicted ldquopossibility of diseaserdquo ge 0239 68 72 139 70 0788

MS + past caries Predicted ldquopossibility of diseaserdquo ge0380 81 77 158 79 0841II

LB + past caries Predicted ldquopossibility of diseaserdquo ge0406 75 81 156 78 0805II

MS + LB + past caries Predicted ldquopossibility of diseaserdquo ge0435 80 80 160 80 0855III

Multifactorial modelClinical screening model Predicted ldquopossibility of diseaserdquo ge0328 82 73 155 77 0845II

Without past caries Predicted ldquopossibility of diseaserdquo ge0416 75 76 151 75 0803II

Multifactorial model (MS) Predicted ldquopossibility of diseaserdquo ge0362 85 78 164 81 0888III

Without past caries Predicted ldquopossibility of diseaserdquo ge0385 85 77 162 81 0875III

Multifactorial model (LB) Predicted ldquopossibility of diseaserdquo ge0434 78 85 163 82 0862III

Without past caries Predicted ldquopossibility of diseaserdquo ge0324 83 72 155 77 0850III

Multifactorial model (MS + LB)

Predicted ldquopossibility of diseaserdquo ge0352 81 85 166 83 0897I

Without past caries Predicted ldquopossibility of diseaserdquo ge0280 85 80 165 82 0889III

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 353

LB level reached ldquoScore 2rdquo It is possible that an increased LB count includes an increase in the level of probiotic LB species balancing out the effect of cariogenic species Such an assump- tion is supported by studies showing that many Lactobacillus species in saliva significantly inhibit the growth of MS2526 Nevertheless since our study did not include a quantification of different LB species this assumption is yet to be tested

It has been suggested that LB may be a better caries pre- dictor compared with MS count27 Our study showed no signifi- cant difference in the ldquosensitivity plus specificityrdquo value between MS (146 percent) and LB (140 percent) However MS achieved a significantly higher sensitivity (79 percent vs 51 percent) whereas LB showed a higher specificity (89 percent vs 67 percent) This substantiated previously published data collected from preschoolers2829 Adding LB to MS without a third factor involved appeared to be of little value if one considers the similar Se+Sp (146 percent vs 151 percent) of MS model and MS+LB model and the additional cost spent on the LB test Among children who were caries free at baseline the combina- tion of MS and LB could only generate a SeSp of 68 percent 72 percent showing limited potential of these two bacterial tests in identifying caries-free children for early prevention

An ldquoMS+past cariesrdquo model has been chosen by some re- searchers as a simple screening tool for selecting children with high caries risk3031 However our findings suggested that adding MS to ldquopast cariesrdquo did not significantly improve accuracy (from 77-79 percent) or AUC Only after the addition of both MS and LB levels did the prediction accuracy of ldquopast cariesrdquo im- prove with a sensitivityspecificity of 80 percent80 percent and a significantly higher AUC (0855 versus 0783 Plt05)

Role of microbiological factors in multifactorial models It is encouraging that the nonmicrobiological clinical screening model (sensitivityspecificity of 82 percent73 percent) achieved a similar performance to the model combining microbiological tests and ldquopast cariesrdquo (80 percent80 percent) This highlights the importance of proper history taking and oral examination in identifying at-risk preschoolers Incorporating either MS or LB tests into the multifactorial model elevated the summation of sensitivity and specificity to above 160 percent with no sig- nificant improvement in AUC Only when both MS and LB tests were incorporated was the AUC significantly increased (0889 versus 0845 Plt05) with balanced sensitivity (85 percent) and specificity (80 percent) The incorporation of MS and LB tests into the multifactorial model mainly improved the Sp (73-85 percent) with minimal changes in Se (82-81 per- cent) This improvement in specificity is of particular impor- tance for countries with relatively low caries rates where false positive diagnosis may thus incur considerable overtreatment and waste of resources

A recent study showed a sharp decrease in specificity (from 60 percent to 47 percent) when MS was excluded from a multi- factorial CRA program called Cariogram32 However such a drastic decrease did not appear in our data for the multifac- torial model (sensitivityspecificity 85 percent78 percent with MS versus 82 percent73 percent without MS) seemingly con- firming the critical contribution of sociodemographic factors in our models and the related weighting system established through our relatively large-scale longitudinal data13

When ldquopast cariesrdquo was excluded from the modeling multi-factorial models involving MS still predicted ECC adequately reaching ldquosensitivity plus specificityrdquo values of 165 percent and 162 percent with and without LB respectively This result im- plies that MS and ldquopast cariesrdquo may be playing a similar role in

the multifactorial model involving sociodemographic behav-ioral and clinical factors Although ldquopast cariesrdquo is regarded as the best indicator for future caries this variable actually reflects the ldquoconsequencerdquo and not the ldquocauserdquo of caries Replacing ldquopast cariesrdquo with a modifiable MS factor may yield a multifacto-rial model that could be effectively utilized for patient or parent education monitoring disease activity and motivation building during follow-up visits Although the current cost to quantify MS and LB is not inexpensive the advancing technologies might substantially reduce the overall cost when the applica- tion of these tests becomes widely accepted and the commercial value is demonstrated as has happened with other health care products and services

This studyrsquos findings were generated from a child popula- tion fully covered by a fluoridated public water supply Further investigations may be needed for understanding the role of microbiological factors in predicting ECC in nonfluoridated communities

ConclusionsBased on the findings of this study the following conclusions can be drawn

1 The combined lsquomutans streptococci + lactobacilli + past cariesrsquo model with both sensitivity and specifi- city of 80 percent may serve as a useful method for selecting at-risk children for targeted intervention

2 Proper history taking and oral examination remain pertinent in identifying at risk preschoolers especially in rural or financially disadvantaged communities

3 When resources allow incorporating MS and LB into a biopsychosociobehavioral model may offer accurate prediction of ECC among preschoolers even without lsquopast cariesrsquo status

AcknowledgmentsThe authors wish to thank the participating kindergartens for their kind support Dr Robert Yee for his valuable comments on the manuscript and Dr Trevor Lane for his editorial assist- ance This study was financially supported by the Singapore Ministry of Education Academic Research Funds R222-000-021-112 and R222-000-022-112 The funders had no role in study design data collection and analysis decision to publish or preparation of the manuscript

References1 World Health Organization Oral Health CountryArea

Profile Programme (CAPP) Geneva WHO 2006 Avail- able at ldquohttpwwwmahsecapprdquo Accessed June 12 2014

2 US Department of Health and Human Services Trend in Oral Health Status United States 1988-1994 And 1999-2004 Atlanta Ga Centers for Disease Control and Prevention National Center for Health Statistics 2007

3 Pitts NB Chestnutt IG Evans D White D Chadwick B Steele JG The dentinal caries experience of children in the United Kingdom 2003 Br Dent J 200625313-20

4 Speechley M Johnston DW Some evidence from Ontario Canada of a reversal in the dental caries decline Caries Res 199630423-7

5 Armfield JM Roberts-Thomson KF Slade GD Spencer AJ Dental Health Differences between Boys and Girls The Child Dental Health Survey Australia 2000 AIHW cat no DEN 131 Canberra Australia Australian Institute

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105

350 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

were exclusively restoration or extraction No preventive re- gimens such as fluoride applications sealants or antimicrobial therapies were recorded

At baseline 718 (40 percent) children were affected by caries The mean (plusmnSD) dmft were 154 (plusmn275) The majority (90 percent) of the affected teeth were untreated decayed teeth In 12 months 688 (44 percent) children developed new carious lesions (dmft increment Δdmftgt0 Table 1) including 206 children who were cavity-free at baseline and 482 children who had cavities at baseline and developed a new cavity on at least one originally intact tooth High MS levels (Score 2 or 3 ge105 CFUml saliva)24 and high LB levels (score 1-3 ge104 CFUml saliva)16 were found in 53 percent and 28 percent of children respectively

Associations between each predictive factor and new caries (dmft increment Δdmftgt0) were expressed as relative risk (RR) after controlling for sociodemographic and behavioral factors (Table 1) Compared with children having an MS score of 0 those with scores of 1 2 and 3 were significantly more likely to have new caries as reflected by an RR (95 CI) of 196

(126-283) 341 (259-419) and 464 (389-523) respectively Compared with children whose LB score was 0 those scored as 1 2 and 3 had an RR (95 CI) of 193 (144-237) 272 (221-302) and 270 (223-299) respectively Compared with children who were cavity-free at baseline those found with any cavity at the start of the study (baseline dmftgt0) were 162 (95 CI=125-190) times as likely to develop caries in 12 months Among children who were caries free at baseline the mean caries increment percent with new caries and RR for each stratum was lower vs statistics from the whole sample

Multifactorial CRA models involving MS LB and both are described in Table 2 Other variables included in one or more of these models were age ethnicity parental educa-tion breast-feeding history intake of sweets use of fluoride- containing toothpaste systemic health past caries and oral hygiene status The performances of these multifactorial models single factors (MS LB and ldquopast cariesrdquo) and their combinations in assessing childrenrsquos caries risk are listed and compared in Table 3 The optimal cutoff point of the MS level for predicting caries was ldquoCFUml saliva ge105 (score 2)rdquo

Table 1 CARIES INCREMENT IN CHILDREN WITH DIFFERENT PAST CARIES EXPERIENCE AND DIFFERENT BACTERIAL LEVELS

Factors n Mean (SD)caries increment

(Δdmft)

withnew caries(Δdmftgt0)

RR (95 CI)dagger P-value

MS level (all children) Dentocult score 0 (lt104 CFUml saliva) 533 34 030 (084)I 16 1 (reference) lt001

Dentocult score 1 (104 -105 CFUml saliva)

198 13 058 (131)I 28 196 (126-283)

Dentocult score 2 (105 -106 CFUml saliva)

382 25 111 (150)II 53 341 (259-419)

Dentocult score 3 (gt106 CFUml saliva) 445 29 165 (153)III 75 464 (389-523)

MS level (children with baseline dmft=0)

Dentocult score 0 (lt104 CFUml saliva) 464 50 022 (073) 110 1 (reference)

Dentocult score 1 (104 -105 CFUml saliva)

143 15 038 (121) 182 177 (129-206)

Dentocult score 2 (105 -106 CFUml saliva)

195 21 056 (113) 282 200 (172-216)

Dentocult score 3 (gt106 CFUml saliva) 133 14 105 (138) 534 219 (204-225)

LB level (all children) Dentocult score 0 (103 CFUml saliva) 1029 72 061 (123)a 30 1 (reference) lt001

Dentocult score 1 (104 CFUml saliva) 148 10 130 (132)b 69 193 (144-237)

Dentocult score 2 (105 CFUml saliva) 99 7 194 (177)c 81 272 (221-302)

Dentocult score 3 (106 CFUml saliva) 160 11 201 (141)c 88 270 (223-299)

LB level (children with baseline dmft=0)

Dentocult score 0 (103 CFUml saliva) 779 92 038 (102) 19 1 (reference)

Dentocult score 1 (104 CFUml saliva) 31 4 058 (085) 39 166 (070-215)

Dentocult score 2 (105 CFUml saliva) 26 3 162 (158) 65 202 (056-228)

Dentocult score 3 (106 CFUml saliva) 14 2 186 (117) 79 216 (156-227)

Past cariesDagger Cavity-free (baseline dmft=0) 940 60 044 (105)i 22 1 (reference) lt001

With cavity (baseline dmftgt0) 636 40 166 (157)ii 76 162 (125-190)

Total 1576 100 093 (144) 44

Ranking (I-III i-ii and a-c) there was significant difference in mean caries increment (∆dmft) between strata with different ranks Tukeyrsquos post-hoc tests or independent t tests (as appropriate) were used to compare means while Kruskal-Wallis tests or Mann-Whitney tests (as appropriate) were employed when the normal distribution or homo- geneity of variance was violateddagger Odds Ratios (OR) and their 95 confidence intervals (CI) were derived from multiple logistic regressions after adjusting for demographic socioeconomic and behavioral factors For a better reflection of the risk in this prospective cohort study ORs were converted to relative risk (RR) thorough the following equation RR=OR[(1-P0) + (P0 X OR)] (P0 stands for the disease incidence)Dagger Both microbiological factors were benchmarked against ldquopast cariesrdquo which is regarded as the strongest caries indicator9

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 351

reaching a sensitivityspecificity of 79 percent67 percent For the LB level the optimal cutoff point was score 1 (ge104 CFUml saliva) with a sensitivityspecificity of 51 percent89 percent in predicting caries When MS was combined with LB the sensitivityspecificity (66 percent85 percent) was similar to that of ldquopast cariesrdquo (70 percent83 percent) Adding LB or MS to ldquopast cariesrdquo minimally increased its sen- sitivity or specificity When both microbiological factors (MS and LB) were combined with ldquopast cariesrdquo the sensitivity and specificity totaled 160 percent with each being 80 percent

The nonmicrobiological clinical screening model that included only psychobehavioral questionnaire in- formation and clinical examination results (oral hygiene status and ldquopast cariesrdquo) yielded an AUC close to that of the combined ldquoMS+LB+past cariesrdquo model (0845 vs 0855) However its specificity was slightly lower (73 percent vs 80 percent) Incorporating only the LB test into that clinical screening model improved the spe- cificity considerably (73-85 percent) with a slight decrease in sensitivity (82-78 percent) When both MS and LB were added to the multifactorial model signifi- cantly higher overall performances (AUC 0845-0897) were observed with a sensitivityspecificity of 81 percent 85 percent Subtraction of ldquopast cariesrdquo from the mutifac- torial models revealed that this factor contributed mini- mally in the multifactorial models involving MS (alone or with LB) Without ldquopast cariesrdquo the multifactorial model involving MS and LB reached a sensitivity specificity of 85 percent80 percent The ROC curves of combinations of factors and multifactorial models are presented in Figure 1 paralleling the aforementioned findings

In the subgroup of children who were caries free at baseline the low specificity of MS level and low sensiti- vity of LB level in predicting caries was further exacer- bated When both tests were combined the SeSp was 68 percent72 percent with a summation of 139 percent Compared with the corresponding statistics in the whole sample (SeSp=66 percent85 percent) a drop in specificity was observed indicating more false positive cases in this sub- group of children

DiscussionMethodological considerations This popula-tion-based prospective study in a relatively large sample aimed to generate new evidence on the usefulness of microbiological parameters in predicting ECC as single factors in combina- tion and in multifactorial modeling A limita- tion of this study was that initial noncavitated carious lesions were not registered The findings of this study are therefore better understood as appraisal of caries prediction methods at the cavitation level Although 62 (four percent) children had visited a dentist during the study period the types of dental care they had re- ceived were exclusively curative (restoration and extraction) without any preventive measures Therefore the confounding impacts of these dental visits may be negligible Commercially

available diagnostic kits were used to evaluate the level of cario- genic bacteria in this study Although their accuracy may be less ideal compared with advanced laboratory technologies

Odds ratios and their 95 confidence intervals (CI) were derived logistic regressions The out- come variable was ldquoany caries incrementrdquo in a year (∆dmft gt0 or =0)

Table 2 MULTIFACTORIAL CARIES PREDICTION MODELS INVOLVING MUTANS STREPTOCOCCI (MS) ANDOR LACTOBACILLI (LB)

Odds ratio (95 CI)

Multifactorial model with MS

Multifactorial model with LB

Multifactorial model with MS

and LB

Questionnaire informationAge (mos) 1056 (1029-1084) 1041 (1015-1067) 1051 (1023-1078)

Malay race 1764 (1096-2841)

Fatherrsquos education

0668 (0552-0809) 0630 (0533-0744) 0636 (0527-0738)

Months of breastfeeding

1045 (1017-1074) 1040 (1016-1065) 1059 (1030-1087)

Frequency of sweet

1393 (1112-1746)

Fluoridated toothpaste

0580 (0346-0973) 0554 (0340-0903) 0573 (0343-0953)

No health problems

2436 (1428-4156) 2352 (1451-3815) 2153 (1263-3618)

Clinical dataPast (baseline) caries

4279 (2899-6314) 4820 (3285-7073) 2983 (1698-5497)

Plaque index 5205 (3227-8394) 5212 (3350-8109) 5197(3220-8385)

Biological testsMS 2161 (1839-2539) 2104 (1782-2486)

LB 1932 (1549-2409) 1859 (1161-2998)

Figure 1 Receiver operation characteristics (ROC) curves for models involving two or more factors

352 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

The performance of our previously reported clinical screening model and past caries which is regarded as the strongest predictor for future caries is pre- sented in this table for comparison The clinical screening model requires information on sociodemographic background oral health behaviors and clinical conditions (oral hygiene and past caries)13

dagger Receiver operation characteristics (ROC) curves were only plotted when there are multiple possible cutoff points Area under ROC curve (AUC) were there- fore only available for these modelsDagger Cutoff point with highest ldquosensitivity (Se) + specificity (Sp)rdquo in indicating caries Data under these optimal cutoff points were bolded for easy comparisonsRanking (I-II) There was a significant difference in AUC between factorscombinationsmodels with different ranks

(immunoassay and agar-based quantification) they are validated and easy to use require little apparatus and are currently the only common and practical option in the clinical setting

Dental caries appeared to be more prevalent in Singaporean children than among children in Western countries This is in line with the ECC pattern in Asian populations1 The skewed distribution of caries lesions in this sample however justifies the

need for identifying high-risk children for intensive prevention and intervention

Single microbiological factors and their combinations As a single factor both MS and LB achieved an accuracy of 72 percent in predicting ECCmdashlower than that of ldquopast cariesrdquo (77 percent) Although there was a positive gradient between MS level and ECC caries risk approached a plateau when the

Table 3 PREDICTIVE VALUES USING SINGLE FACTORS (MUTANS STREPTOCOCCI [MS] LACTOBACILLI [LB] AND ldquoPAST CARIESrdquo) THEIR COMBINATIONS AND MULTIFACTORIAL MODELS

Screening criteria for children with any risk (Δdmftgt0)

Performance measures

AUCdagger

Se Sp Se+Sp Accuracy

Single factorsPast caries Baseline dmft gt0 70 83 153 77 NA

MS level (all children) Dentocult score ge1 (ge104 CFUml saliva) 88 51 139 67 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 79 67 146 72

Dentocult score ge3 (gt106 CFUml saliva) 50 88 137 71

MS level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva) 77 43 120 58 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 64 27 91 43

Dentocult score ge3 (gt106 CFUml saliva) 35 7 42 19

LB level (all children) Dentocult score ge1 (ge104 CFUml saliva)Dagger 51 89 140 72 NA

Dentocult score ge2 (ge105 CFUml saliva) 35 95 130 69

Dentocult score ge3 (gt106 CFUml saliva) 22 98 120 65

LB level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva)Dagger 21 95 117 63 NA

Dentocult score ge2 (ge105 CFUml saliva) 15 98 113 62

Dentocult score ge3 (gt106 CFUml saliva) 6 99 105 58

CombinationsMS + LB (all children) Predicted ldquopossibility of diseaserdquo ge0524 66 85 151 77 0822II

MS + LB (children with baseline dmft=0)

Predicted ldquopossibility of diseaserdquo ge 0239 68 72 139 70 0788

MS + past caries Predicted ldquopossibility of diseaserdquo ge0380 81 77 158 79 0841II

LB + past caries Predicted ldquopossibility of diseaserdquo ge0406 75 81 156 78 0805II

MS + LB + past caries Predicted ldquopossibility of diseaserdquo ge0435 80 80 160 80 0855III

Multifactorial modelClinical screening model Predicted ldquopossibility of diseaserdquo ge0328 82 73 155 77 0845II

Without past caries Predicted ldquopossibility of diseaserdquo ge0416 75 76 151 75 0803II

Multifactorial model (MS) Predicted ldquopossibility of diseaserdquo ge0362 85 78 164 81 0888III

Without past caries Predicted ldquopossibility of diseaserdquo ge0385 85 77 162 81 0875III

Multifactorial model (LB) Predicted ldquopossibility of diseaserdquo ge0434 78 85 163 82 0862III

Without past caries Predicted ldquopossibility of diseaserdquo ge0324 83 72 155 77 0850III

Multifactorial model (MS + LB)

Predicted ldquopossibility of diseaserdquo ge0352 81 85 166 83 0897I

Without past caries Predicted ldquopossibility of diseaserdquo ge0280 85 80 165 82 0889III

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 353

LB level reached ldquoScore 2rdquo It is possible that an increased LB count includes an increase in the level of probiotic LB species balancing out the effect of cariogenic species Such an assump- tion is supported by studies showing that many Lactobacillus species in saliva significantly inhibit the growth of MS2526 Nevertheless since our study did not include a quantification of different LB species this assumption is yet to be tested

It has been suggested that LB may be a better caries pre- dictor compared with MS count27 Our study showed no signifi- cant difference in the ldquosensitivity plus specificityrdquo value between MS (146 percent) and LB (140 percent) However MS achieved a significantly higher sensitivity (79 percent vs 51 percent) whereas LB showed a higher specificity (89 percent vs 67 percent) This substantiated previously published data collected from preschoolers2829 Adding LB to MS without a third factor involved appeared to be of little value if one considers the similar Se+Sp (146 percent vs 151 percent) of MS model and MS+LB model and the additional cost spent on the LB test Among children who were caries free at baseline the combina- tion of MS and LB could only generate a SeSp of 68 percent 72 percent showing limited potential of these two bacterial tests in identifying caries-free children for early prevention

An ldquoMS+past cariesrdquo model has been chosen by some re- searchers as a simple screening tool for selecting children with high caries risk3031 However our findings suggested that adding MS to ldquopast cariesrdquo did not significantly improve accuracy (from 77-79 percent) or AUC Only after the addition of both MS and LB levels did the prediction accuracy of ldquopast cariesrdquo im- prove with a sensitivityspecificity of 80 percent80 percent and a significantly higher AUC (0855 versus 0783 Plt05)

Role of microbiological factors in multifactorial models It is encouraging that the nonmicrobiological clinical screening model (sensitivityspecificity of 82 percent73 percent) achieved a similar performance to the model combining microbiological tests and ldquopast cariesrdquo (80 percent80 percent) This highlights the importance of proper history taking and oral examination in identifying at-risk preschoolers Incorporating either MS or LB tests into the multifactorial model elevated the summation of sensitivity and specificity to above 160 percent with no sig- nificant improvement in AUC Only when both MS and LB tests were incorporated was the AUC significantly increased (0889 versus 0845 Plt05) with balanced sensitivity (85 percent) and specificity (80 percent) The incorporation of MS and LB tests into the multifactorial model mainly improved the Sp (73-85 percent) with minimal changes in Se (82-81 per- cent) This improvement in specificity is of particular impor- tance for countries with relatively low caries rates where false positive diagnosis may thus incur considerable overtreatment and waste of resources

A recent study showed a sharp decrease in specificity (from 60 percent to 47 percent) when MS was excluded from a multi- factorial CRA program called Cariogram32 However such a drastic decrease did not appear in our data for the multifac- torial model (sensitivityspecificity 85 percent78 percent with MS versus 82 percent73 percent without MS) seemingly con- firming the critical contribution of sociodemographic factors in our models and the related weighting system established through our relatively large-scale longitudinal data13

When ldquopast cariesrdquo was excluded from the modeling multi-factorial models involving MS still predicted ECC adequately reaching ldquosensitivity plus specificityrdquo values of 165 percent and 162 percent with and without LB respectively This result im- plies that MS and ldquopast cariesrdquo may be playing a similar role in

the multifactorial model involving sociodemographic behav-ioral and clinical factors Although ldquopast cariesrdquo is regarded as the best indicator for future caries this variable actually reflects the ldquoconsequencerdquo and not the ldquocauserdquo of caries Replacing ldquopast cariesrdquo with a modifiable MS factor may yield a multifacto-rial model that could be effectively utilized for patient or parent education monitoring disease activity and motivation building during follow-up visits Although the current cost to quantify MS and LB is not inexpensive the advancing technologies might substantially reduce the overall cost when the applica- tion of these tests becomes widely accepted and the commercial value is demonstrated as has happened with other health care products and services

This studyrsquos findings were generated from a child popula- tion fully covered by a fluoridated public water supply Further investigations may be needed for understanding the role of microbiological factors in predicting ECC in nonfluoridated communities

ConclusionsBased on the findings of this study the following conclusions can be drawn

1 The combined lsquomutans streptococci + lactobacilli + past cariesrsquo model with both sensitivity and specifi- city of 80 percent may serve as a useful method for selecting at-risk children for targeted intervention

2 Proper history taking and oral examination remain pertinent in identifying at risk preschoolers especially in rural or financially disadvantaged communities

3 When resources allow incorporating MS and LB into a biopsychosociobehavioral model may offer accurate prediction of ECC among preschoolers even without lsquopast cariesrsquo status

AcknowledgmentsThe authors wish to thank the participating kindergartens for their kind support Dr Robert Yee for his valuable comments on the manuscript and Dr Trevor Lane for his editorial assist- ance This study was financially supported by the Singapore Ministry of Education Academic Research Funds R222-000-021-112 and R222-000-022-112 The funders had no role in study design data collection and analysis decision to publish or preparation of the manuscript

References1 World Health Organization Oral Health CountryArea

Profile Programme (CAPP) Geneva WHO 2006 Avail- able at ldquohttpwwwmahsecapprdquo Accessed June 12 2014

2 US Department of Health and Human Services Trend in Oral Health Status United States 1988-1994 And 1999-2004 Atlanta Ga Centers for Disease Control and Prevention National Center for Health Statistics 2007

3 Pitts NB Chestnutt IG Evans D White D Chadwick B Steele JG The dentinal caries experience of children in the United Kingdom 2003 Br Dent J 200625313-20

4 Speechley M Johnston DW Some evidence from Ontario Canada of a reversal in the dental caries decline Caries Res 199630423-7

5 Armfield JM Roberts-Thomson KF Slade GD Spencer AJ Dental Health Differences between Boys and Girls The Child Dental Health Survey Australia 2000 AIHW cat no DEN 131 Canberra Australia Australian Institute

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 351

reaching a sensitivityspecificity of 79 percent67 percent For the LB level the optimal cutoff point was score 1 (ge104 CFUml saliva) with a sensitivityspecificity of 51 percent89 percent in predicting caries When MS was combined with LB the sensitivityspecificity (66 percent85 percent) was similar to that of ldquopast cariesrdquo (70 percent83 percent) Adding LB or MS to ldquopast cariesrdquo minimally increased its sen- sitivity or specificity When both microbiological factors (MS and LB) were combined with ldquopast cariesrdquo the sensitivity and specificity totaled 160 percent with each being 80 percent

The nonmicrobiological clinical screening model that included only psychobehavioral questionnaire in- formation and clinical examination results (oral hygiene status and ldquopast cariesrdquo) yielded an AUC close to that of the combined ldquoMS+LB+past cariesrdquo model (0845 vs 0855) However its specificity was slightly lower (73 percent vs 80 percent) Incorporating only the LB test into that clinical screening model improved the spe- cificity considerably (73-85 percent) with a slight decrease in sensitivity (82-78 percent) When both MS and LB were added to the multifactorial model signifi- cantly higher overall performances (AUC 0845-0897) were observed with a sensitivityspecificity of 81 percent 85 percent Subtraction of ldquopast cariesrdquo from the mutifac- torial models revealed that this factor contributed mini- mally in the multifactorial models involving MS (alone or with LB) Without ldquopast cariesrdquo the multifactorial model involving MS and LB reached a sensitivity specificity of 85 percent80 percent The ROC curves of combinations of factors and multifactorial models are presented in Figure 1 paralleling the aforementioned findings

In the subgroup of children who were caries free at baseline the low specificity of MS level and low sensiti- vity of LB level in predicting caries was further exacer- bated When both tests were combined the SeSp was 68 percent72 percent with a summation of 139 percent Compared with the corresponding statistics in the whole sample (SeSp=66 percent85 percent) a drop in specificity was observed indicating more false positive cases in this sub- group of children

DiscussionMethodological considerations This popula-tion-based prospective study in a relatively large sample aimed to generate new evidence on the usefulness of microbiological parameters in predicting ECC as single factors in combina- tion and in multifactorial modeling A limita- tion of this study was that initial noncavitated carious lesions were not registered The findings of this study are therefore better understood as appraisal of caries prediction methods at the cavitation level Although 62 (four percent) children had visited a dentist during the study period the types of dental care they had re- ceived were exclusively curative (restoration and extraction) without any preventive measures Therefore the confounding impacts of these dental visits may be negligible Commercially

available diagnostic kits were used to evaluate the level of cario- genic bacteria in this study Although their accuracy may be less ideal compared with advanced laboratory technologies

Odds ratios and their 95 confidence intervals (CI) were derived logistic regressions The out- come variable was ldquoany caries incrementrdquo in a year (∆dmft gt0 or =0)

Table 2 MULTIFACTORIAL CARIES PREDICTION MODELS INVOLVING MUTANS STREPTOCOCCI (MS) ANDOR LACTOBACILLI (LB)

Odds ratio (95 CI)

Multifactorial model with MS

Multifactorial model with LB

Multifactorial model with MS

and LB

Questionnaire informationAge (mos) 1056 (1029-1084) 1041 (1015-1067) 1051 (1023-1078)

Malay race 1764 (1096-2841)

Fatherrsquos education

0668 (0552-0809) 0630 (0533-0744) 0636 (0527-0738)

Months of breastfeeding

1045 (1017-1074) 1040 (1016-1065) 1059 (1030-1087)

Frequency of sweet

1393 (1112-1746)

Fluoridated toothpaste

0580 (0346-0973) 0554 (0340-0903) 0573 (0343-0953)

No health problems

2436 (1428-4156) 2352 (1451-3815) 2153 (1263-3618)

Clinical dataPast (baseline) caries

4279 (2899-6314) 4820 (3285-7073) 2983 (1698-5497)

Plaque index 5205 (3227-8394) 5212 (3350-8109) 5197(3220-8385)

Biological testsMS 2161 (1839-2539) 2104 (1782-2486)

LB 1932 (1549-2409) 1859 (1161-2998)

Figure 1 Receiver operation characteristics (ROC) curves for models involving two or more factors

352 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

The performance of our previously reported clinical screening model and past caries which is regarded as the strongest predictor for future caries is pre- sented in this table for comparison The clinical screening model requires information on sociodemographic background oral health behaviors and clinical conditions (oral hygiene and past caries)13

dagger Receiver operation characteristics (ROC) curves were only plotted when there are multiple possible cutoff points Area under ROC curve (AUC) were there- fore only available for these modelsDagger Cutoff point with highest ldquosensitivity (Se) + specificity (Sp)rdquo in indicating caries Data under these optimal cutoff points were bolded for easy comparisonsRanking (I-II) There was a significant difference in AUC between factorscombinationsmodels with different ranks

(immunoassay and agar-based quantification) they are validated and easy to use require little apparatus and are currently the only common and practical option in the clinical setting

Dental caries appeared to be more prevalent in Singaporean children than among children in Western countries This is in line with the ECC pattern in Asian populations1 The skewed distribution of caries lesions in this sample however justifies the

need for identifying high-risk children for intensive prevention and intervention

Single microbiological factors and their combinations As a single factor both MS and LB achieved an accuracy of 72 percent in predicting ECCmdashlower than that of ldquopast cariesrdquo (77 percent) Although there was a positive gradient between MS level and ECC caries risk approached a plateau when the

Table 3 PREDICTIVE VALUES USING SINGLE FACTORS (MUTANS STREPTOCOCCI [MS] LACTOBACILLI [LB] AND ldquoPAST CARIESrdquo) THEIR COMBINATIONS AND MULTIFACTORIAL MODELS

Screening criteria for children with any risk (Δdmftgt0)

Performance measures

AUCdagger

Se Sp Se+Sp Accuracy

Single factorsPast caries Baseline dmft gt0 70 83 153 77 NA

MS level (all children) Dentocult score ge1 (ge104 CFUml saliva) 88 51 139 67 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 79 67 146 72

Dentocult score ge3 (gt106 CFUml saliva) 50 88 137 71

MS level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva) 77 43 120 58 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 64 27 91 43

Dentocult score ge3 (gt106 CFUml saliva) 35 7 42 19

LB level (all children) Dentocult score ge1 (ge104 CFUml saliva)Dagger 51 89 140 72 NA

Dentocult score ge2 (ge105 CFUml saliva) 35 95 130 69

Dentocult score ge3 (gt106 CFUml saliva) 22 98 120 65

LB level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva)Dagger 21 95 117 63 NA

Dentocult score ge2 (ge105 CFUml saliva) 15 98 113 62

Dentocult score ge3 (gt106 CFUml saliva) 6 99 105 58

CombinationsMS + LB (all children) Predicted ldquopossibility of diseaserdquo ge0524 66 85 151 77 0822II

MS + LB (children with baseline dmft=0)

Predicted ldquopossibility of diseaserdquo ge 0239 68 72 139 70 0788

MS + past caries Predicted ldquopossibility of diseaserdquo ge0380 81 77 158 79 0841II

LB + past caries Predicted ldquopossibility of diseaserdquo ge0406 75 81 156 78 0805II

MS + LB + past caries Predicted ldquopossibility of diseaserdquo ge0435 80 80 160 80 0855III

Multifactorial modelClinical screening model Predicted ldquopossibility of diseaserdquo ge0328 82 73 155 77 0845II

Without past caries Predicted ldquopossibility of diseaserdquo ge0416 75 76 151 75 0803II

Multifactorial model (MS) Predicted ldquopossibility of diseaserdquo ge0362 85 78 164 81 0888III

Without past caries Predicted ldquopossibility of diseaserdquo ge0385 85 77 162 81 0875III

Multifactorial model (LB) Predicted ldquopossibility of diseaserdquo ge0434 78 85 163 82 0862III

Without past caries Predicted ldquopossibility of diseaserdquo ge0324 83 72 155 77 0850III

Multifactorial model (MS + LB)

Predicted ldquopossibility of diseaserdquo ge0352 81 85 166 83 0897I

Without past caries Predicted ldquopossibility of diseaserdquo ge0280 85 80 165 82 0889III

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 353

LB level reached ldquoScore 2rdquo It is possible that an increased LB count includes an increase in the level of probiotic LB species balancing out the effect of cariogenic species Such an assump- tion is supported by studies showing that many Lactobacillus species in saliva significantly inhibit the growth of MS2526 Nevertheless since our study did not include a quantification of different LB species this assumption is yet to be tested

It has been suggested that LB may be a better caries pre- dictor compared with MS count27 Our study showed no signifi- cant difference in the ldquosensitivity plus specificityrdquo value between MS (146 percent) and LB (140 percent) However MS achieved a significantly higher sensitivity (79 percent vs 51 percent) whereas LB showed a higher specificity (89 percent vs 67 percent) This substantiated previously published data collected from preschoolers2829 Adding LB to MS without a third factor involved appeared to be of little value if one considers the similar Se+Sp (146 percent vs 151 percent) of MS model and MS+LB model and the additional cost spent on the LB test Among children who were caries free at baseline the combina- tion of MS and LB could only generate a SeSp of 68 percent 72 percent showing limited potential of these two bacterial tests in identifying caries-free children for early prevention

An ldquoMS+past cariesrdquo model has been chosen by some re- searchers as a simple screening tool for selecting children with high caries risk3031 However our findings suggested that adding MS to ldquopast cariesrdquo did not significantly improve accuracy (from 77-79 percent) or AUC Only after the addition of both MS and LB levels did the prediction accuracy of ldquopast cariesrdquo im- prove with a sensitivityspecificity of 80 percent80 percent and a significantly higher AUC (0855 versus 0783 Plt05)

Role of microbiological factors in multifactorial models It is encouraging that the nonmicrobiological clinical screening model (sensitivityspecificity of 82 percent73 percent) achieved a similar performance to the model combining microbiological tests and ldquopast cariesrdquo (80 percent80 percent) This highlights the importance of proper history taking and oral examination in identifying at-risk preschoolers Incorporating either MS or LB tests into the multifactorial model elevated the summation of sensitivity and specificity to above 160 percent with no sig- nificant improvement in AUC Only when both MS and LB tests were incorporated was the AUC significantly increased (0889 versus 0845 Plt05) with balanced sensitivity (85 percent) and specificity (80 percent) The incorporation of MS and LB tests into the multifactorial model mainly improved the Sp (73-85 percent) with minimal changes in Se (82-81 per- cent) This improvement in specificity is of particular impor- tance for countries with relatively low caries rates where false positive diagnosis may thus incur considerable overtreatment and waste of resources

A recent study showed a sharp decrease in specificity (from 60 percent to 47 percent) when MS was excluded from a multi- factorial CRA program called Cariogram32 However such a drastic decrease did not appear in our data for the multifac- torial model (sensitivityspecificity 85 percent78 percent with MS versus 82 percent73 percent without MS) seemingly con- firming the critical contribution of sociodemographic factors in our models and the related weighting system established through our relatively large-scale longitudinal data13

When ldquopast cariesrdquo was excluded from the modeling multi-factorial models involving MS still predicted ECC adequately reaching ldquosensitivity plus specificityrdquo values of 165 percent and 162 percent with and without LB respectively This result im- plies that MS and ldquopast cariesrdquo may be playing a similar role in

the multifactorial model involving sociodemographic behav-ioral and clinical factors Although ldquopast cariesrdquo is regarded as the best indicator for future caries this variable actually reflects the ldquoconsequencerdquo and not the ldquocauserdquo of caries Replacing ldquopast cariesrdquo with a modifiable MS factor may yield a multifacto-rial model that could be effectively utilized for patient or parent education monitoring disease activity and motivation building during follow-up visits Although the current cost to quantify MS and LB is not inexpensive the advancing technologies might substantially reduce the overall cost when the applica- tion of these tests becomes widely accepted and the commercial value is demonstrated as has happened with other health care products and services

This studyrsquos findings were generated from a child popula- tion fully covered by a fluoridated public water supply Further investigations may be needed for understanding the role of microbiological factors in predicting ECC in nonfluoridated communities

ConclusionsBased on the findings of this study the following conclusions can be drawn

1 The combined lsquomutans streptococci + lactobacilli + past cariesrsquo model with both sensitivity and specifi- city of 80 percent may serve as a useful method for selecting at-risk children for targeted intervention

2 Proper history taking and oral examination remain pertinent in identifying at risk preschoolers especially in rural or financially disadvantaged communities

3 When resources allow incorporating MS and LB into a biopsychosociobehavioral model may offer accurate prediction of ECC among preschoolers even without lsquopast cariesrsquo status

AcknowledgmentsThe authors wish to thank the participating kindergartens for their kind support Dr Robert Yee for his valuable comments on the manuscript and Dr Trevor Lane for his editorial assist- ance This study was financially supported by the Singapore Ministry of Education Academic Research Funds R222-000-021-112 and R222-000-022-112 The funders had no role in study design data collection and analysis decision to publish or preparation of the manuscript

References1 World Health Organization Oral Health CountryArea

Profile Programme (CAPP) Geneva WHO 2006 Avail- able at ldquohttpwwwmahsecapprdquo Accessed June 12 2014

2 US Department of Health and Human Services Trend in Oral Health Status United States 1988-1994 And 1999-2004 Atlanta Ga Centers for Disease Control and Prevention National Center for Health Statistics 2007

3 Pitts NB Chestnutt IG Evans D White D Chadwick B Steele JG The dentinal caries experience of children in the United Kingdom 2003 Br Dent J 200625313-20

4 Speechley M Johnston DW Some evidence from Ontario Canada of a reversal in the dental caries decline Caries Res 199630423-7

5 Armfield JM Roberts-Thomson KF Slade GD Spencer AJ Dental Health Differences between Boys and Girls The Child Dental Health Survey Australia 2000 AIHW cat no DEN 131 Canberra Australia Australian Institute

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105

352 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

The performance of our previously reported clinical screening model and past caries which is regarded as the strongest predictor for future caries is pre- sented in this table for comparison The clinical screening model requires information on sociodemographic background oral health behaviors and clinical conditions (oral hygiene and past caries)13

dagger Receiver operation characteristics (ROC) curves were only plotted when there are multiple possible cutoff points Area under ROC curve (AUC) were there- fore only available for these modelsDagger Cutoff point with highest ldquosensitivity (Se) + specificity (Sp)rdquo in indicating caries Data under these optimal cutoff points were bolded for easy comparisonsRanking (I-II) There was a significant difference in AUC between factorscombinationsmodels with different ranks

(immunoassay and agar-based quantification) they are validated and easy to use require little apparatus and are currently the only common and practical option in the clinical setting

Dental caries appeared to be more prevalent in Singaporean children than among children in Western countries This is in line with the ECC pattern in Asian populations1 The skewed distribution of caries lesions in this sample however justifies the

need for identifying high-risk children for intensive prevention and intervention

Single microbiological factors and their combinations As a single factor both MS and LB achieved an accuracy of 72 percent in predicting ECCmdashlower than that of ldquopast cariesrdquo (77 percent) Although there was a positive gradient between MS level and ECC caries risk approached a plateau when the

Table 3 PREDICTIVE VALUES USING SINGLE FACTORS (MUTANS STREPTOCOCCI [MS] LACTOBACILLI [LB] AND ldquoPAST CARIESrdquo) THEIR COMBINATIONS AND MULTIFACTORIAL MODELS

Screening criteria for children with any risk (Δdmftgt0)

Performance measures

AUCdagger

Se Sp Se+Sp Accuracy

Single factorsPast caries Baseline dmft gt0 70 83 153 77 NA

MS level (all children) Dentocult score ge1 (ge104 CFUml saliva) 88 51 139 67 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 79 67 146 72

Dentocult score ge3 (gt106 CFUml saliva) 50 88 137 71

MS level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva) 77 43 120 58 NA

Dentocult score ge2 (ge105 CFUml saliva)Dagger 64 27 91 43

Dentocult score ge3 (gt106 CFUml saliva) 35 7 42 19

LB level (all children) Dentocult score ge1 (ge104 CFUml saliva)Dagger 51 89 140 72 NA

Dentocult score ge2 (ge105 CFUml saliva) 35 95 130 69

Dentocult score ge3 (gt106 CFUml saliva) 22 98 120 65

LB level (children with baseline dmft=0)

Dentocult score ge1 (ge104 CFUml saliva)Dagger 21 95 117 63 NA

Dentocult score ge2 (ge105 CFUml saliva) 15 98 113 62

Dentocult score ge3 (gt106 CFUml saliva) 6 99 105 58

CombinationsMS + LB (all children) Predicted ldquopossibility of diseaserdquo ge0524 66 85 151 77 0822II

MS + LB (children with baseline dmft=0)

Predicted ldquopossibility of diseaserdquo ge 0239 68 72 139 70 0788

MS + past caries Predicted ldquopossibility of diseaserdquo ge0380 81 77 158 79 0841II

LB + past caries Predicted ldquopossibility of diseaserdquo ge0406 75 81 156 78 0805II

MS + LB + past caries Predicted ldquopossibility of diseaserdquo ge0435 80 80 160 80 0855III

Multifactorial modelClinical screening model Predicted ldquopossibility of diseaserdquo ge0328 82 73 155 77 0845II

Without past caries Predicted ldquopossibility of diseaserdquo ge0416 75 76 151 75 0803II

Multifactorial model (MS) Predicted ldquopossibility of diseaserdquo ge0362 85 78 164 81 0888III

Without past caries Predicted ldquopossibility of diseaserdquo ge0385 85 77 162 81 0875III

Multifactorial model (LB) Predicted ldquopossibility of diseaserdquo ge0434 78 85 163 82 0862III

Without past caries Predicted ldquopossibility of diseaserdquo ge0324 83 72 155 77 0850III

Multifactorial model (MS + LB)

Predicted ldquopossibility of diseaserdquo ge0352 81 85 166 83 0897I

Without past caries Predicted ldquopossibility of diseaserdquo ge0280 85 80 165 82 0889III

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 353

LB level reached ldquoScore 2rdquo It is possible that an increased LB count includes an increase in the level of probiotic LB species balancing out the effect of cariogenic species Such an assump- tion is supported by studies showing that many Lactobacillus species in saliva significantly inhibit the growth of MS2526 Nevertheless since our study did not include a quantification of different LB species this assumption is yet to be tested

It has been suggested that LB may be a better caries pre- dictor compared with MS count27 Our study showed no signifi- cant difference in the ldquosensitivity plus specificityrdquo value between MS (146 percent) and LB (140 percent) However MS achieved a significantly higher sensitivity (79 percent vs 51 percent) whereas LB showed a higher specificity (89 percent vs 67 percent) This substantiated previously published data collected from preschoolers2829 Adding LB to MS without a third factor involved appeared to be of little value if one considers the similar Se+Sp (146 percent vs 151 percent) of MS model and MS+LB model and the additional cost spent on the LB test Among children who were caries free at baseline the combina- tion of MS and LB could only generate a SeSp of 68 percent 72 percent showing limited potential of these two bacterial tests in identifying caries-free children for early prevention

An ldquoMS+past cariesrdquo model has been chosen by some re- searchers as a simple screening tool for selecting children with high caries risk3031 However our findings suggested that adding MS to ldquopast cariesrdquo did not significantly improve accuracy (from 77-79 percent) or AUC Only after the addition of both MS and LB levels did the prediction accuracy of ldquopast cariesrdquo im- prove with a sensitivityspecificity of 80 percent80 percent and a significantly higher AUC (0855 versus 0783 Plt05)

Role of microbiological factors in multifactorial models It is encouraging that the nonmicrobiological clinical screening model (sensitivityspecificity of 82 percent73 percent) achieved a similar performance to the model combining microbiological tests and ldquopast cariesrdquo (80 percent80 percent) This highlights the importance of proper history taking and oral examination in identifying at-risk preschoolers Incorporating either MS or LB tests into the multifactorial model elevated the summation of sensitivity and specificity to above 160 percent with no sig- nificant improvement in AUC Only when both MS and LB tests were incorporated was the AUC significantly increased (0889 versus 0845 Plt05) with balanced sensitivity (85 percent) and specificity (80 percent) The incorporation of MS and LB tests into the multifactorial model mainly improved the Sp (73-85 percent) with minimal changes in Se (82-81 per- cent) This improvement in specificity is of particular impor- tance for countries with relatively low caries rates where false positive diagnosis may thus incur considerable overtreatment and waste of resources

A recent study showed a sharp decrease in specificity (from 60 percent to 47 percent) when MS was excluded from a multi- factorial CRA program called Cariogram32 However such a drastic decrease did not appear in our data for the multifac- torial model (sensitivityspecificity 85 percent78 percent with MS versus 82 percent73 percent without MS) seemingly con- firming the critical contribution of sociodemographic factors in our models and the related weighting system established through our relatively large-scale longitudinal data13

When ldquopast cariesrdquo was excluded from the modeling multi-factorial models involving MS still predicted ECC adequately reaching ldquosensitivity plus specificityrdquo values of 165 percent and 162 percent with and without LB respectively This result im- plies that MS and ldquopast cariesrdquo may be playing a similar role in

the multifactorial model involving sociodemographic behav-ioral and clinical factors Although ldquopast cariesrdquo is regarded as the best indicator for future caries this variable actually reflects the ldquoconsequencerdquo and not the ldquocauserdquo of caries Replacing ldquopast cariesrdquo with a modifiable MS factor may yield a multifacto-rial model that could be effectively utilized for patient or parent education monitoring disease activity and motivation building during follow-up visits Although the current cost to quantify MS and LB is not inexpensive the advancing technologies might substantially reduce the overall cost when the applica- tion of these tests becomes widely accepted and the commercial value is demonstrated as has happened with other health care products and services

This studyrsquos findings were generated from a child popula- tion fully covered by a fluoridated public water supply Further investigations may be needed for understanding the role of microbiological factors in predicting ECC in nonfluoridated communities

ConclusionsBased on the findings of this study the following conclusions can be drawn

1 The combined lsquomutans streptococci + lactobacilli + past cariesrsquo model with both sensitivity and specifi- city of 80 percent may serve as a useful method for selecting at-risk children for targeted intervention

2 Proper history taking and oral examination remain pertinent in identifying at risk preschoolers especially in rural or financially disadvantaged communities

3 When resources allow incorporating MS and LB into a biopsychosociobehavioral model may offer accurate prediction of ECC among preschoolers even without lsquopast cariesrsquo status

AcknowledgmentsThe authors wish to thank the participating kindergartens for their kind support Dr Robert Yee for his valuable comments on the manuscript and Dr Trevor Lane for his editorial assist- ance This study was financially supported by the Singapore Ministry of Education Academic Research Funds R222-000-021-112 and R222-000-022-112 The funders had no role in study design data collection and analysis decision to publish or preparation of the manuscript

References1 World Health Organization Oral Health CountryArea

Profile Programme (CAPP) Geneva WHO 2006 Avail- able at ldquohttpwwwmahsecapprdquo Accessed June 12 2014

2 US Department of Health and Human Services Trend in Oral Health Status United States 1988-1994 And 1999-2004 Atlanta Ga Centers for Disease Control and Prevention National Center for Health Statistics 2007

3 Pitts NB Chestnutt IG Evans D White D Chadwick B Steele JG The dentinal caries experience of children in the United Kingdom 2003 Br Dent J 200625313-20

4 Speechley M Johnston DW Some evidence from Ontario Canada of a reversal in the dental caries decline Caries Res 199630423-7

5 Armfield JM Roberts-Thomson KF Slade GD Spencer AJ Dental Health Differences between Boys and Girls The Child Dental Health Survey Australia 2000 AIHW cat no DEN 131 Canberra Australia Australian Institute

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

CARIES PREDICTION USING MICROBIOLOGICAL TESTS 353

LB level reached ldquoScore 2rdquo It is possible that an increased LB count includes an increase in the level of probiotic LB species balancing out the effect of cariogenic species Such an assump- tion is supported by studies showing that many Lactobacillus species in saliva significantly inhibit the growth of MS2526 Nevertheless since our study did not include a quantification of different LB species this assumption is yet to be tested

It has been suggested that LB may be a better caries pre- dictor compared with MS count27 Our study showed no signifi- cant difference in the ldquosensitivity plus specificityrdquo value between MS (146 percent) and LB (140 percent) However MS achieved a significantly higher sensitivity (79 percent vs 51 percent) whereas LB showed a higher specificity (89 percent vs 67 percent) This substantiated previously published data collected from preschoolers2829 Adding LB to MS without a third factor involved appeared to be of little value if one considers the similar Se+Sp (146 percent vs 151 percent) of MS model and MS+LB model and the additional cost spent on the LB test Among children who were caries free at baseline the combina- tion of MS and LB could only generate a SeSp of 68 percent 72 percent showing limited potential of these two bacterial tests in identifying caries-free children for early prevention

An ldquoMS+past cariesrdquo model has been chosen by some re- searchers as a simple screening tool for selecting children with high caries risk3031 However our findings suggested that adding MS to ldquopast cariesrdquo did not significantly improve accuracy (from 77-79 percent) or AUC Only after the addition of both MS and LB levels did the prediction accuracy of ldquopast cariesrdquo im- prove with a sensitivityspecificity of 80 percent80 percent and a significantly higher AUC (0855 versus 0783 Plt05)

Role of microbiological factors in multifactorial models It is encouraging that the nonmicrobiological clinical screening model (sensitivityspecificity of 82 percent73 percent) achieved a similar performance to the model combining microbiological tests and ldquopast cariesrdquo (80 percent80 percent) This highlights the importance of proper history taking and oral examination in identifying at-risk preschoolers Incorporating either MS or LB tests into the multifactorial model elevated the summation of sensitivity and specificity to above 160 percent with no sig- nificant improvement in AUC Only when both MS and LB tests were incorporated was the AUC significantly increased (0889 versus 0845 Plt05) with balanced sensitivity (85 percent) and specificity (80 percent) The incorporation of MS and LB tests into the multifactorial model mainly improved the Sp (73-85 percent) with minimal changes in Se (82-81 per- cent) This improvement in specificity is of particular impor- tance for countries with relatively low caries rates where false positive diagnosis may thus incur considerable overtreatment and waste of resources

A recent study showed a sharp decrease in specificity (from 60 percent to 47 percent) when MS was excluded from a multi- factorial CRA program called Cariogram32 However such a drastic decrease did not appear in our data for the multifac- torial model (sensitivityspecificity 85 percent78 percent with MS versus 82 percent73 percent without MS) seemingly con- firming the critical contribution of sociodemographic factors in our models and the related weighting system established through our relatively large-scale longitudinal data13

When ldquopast cariesrdquo was excluded from the modeling multi-factorial models involving MS still predicted ECC adequately reaching ldquosensitivity plus specificityrdquo values of 165 percent and 162 percent with and without LB respectively This result im- plies that MS and ldquopast cariesrdquo may be playing a similar role in

the multifactorial model involving sociodemographic behav-ioral and clinical factors Although ldquopast cariesrdquo is regarded as the best indicator for future caries this variable actually reflects the ldquoconsequencerdquo and not the ldquocauserdquo of caries Replacing ldquopast cariesrdquo with a modifiable MS factor may yield a multifacto-rial model that could be effectively utilized for patient or parent education monitoring disease activity and motivation building during follow-up visits Although the current cost to quantify MS and LB is not inexpensive the advancing technologies might substantially reduce the overall cost when the applica- tion of these tests becomes widely accepted and the commercial value is demonstrated as has happened with other health care products and services

This studyrsquos findings were generated from a child popula- tion fully covered by a fluoridated public water supply Further investigations may be needed for understanding the role of microbiological factors in predicting ECC in nonfluoridated communities

ConclusionsBased on the findings of this study the following conclusions can be drawn

1 The combined lsquomutans streptococci + lactobacilli + past cariesrsquo model with both sensitivity and specifi- city of 80 percent may serve as a useful method for selecting at-risk children for targeted intervention

2 Proper history taking and oral examination remain pertinent in identifying at risk preschoolers especially in rural or financially disadvantaged communities

3 When resources allow incorporating MS and LB into a biopsychosociobehavioral model may offer accurate prediction of ECC among preschoolers even without lsquopast cariesrsquo status

AcknowledgmentsThe authors wish to thank the participating kindergartens for their kind support Dr Robert Yee for his valuable comments on the manuscript and Dr Trevor Lane for his editorial assist- ance This study was financially supported by the Singapore Ministry of Education Academic Research Funds R222-000-021-112 and R222-000-022-112 The funders had no role in study design data collection and analysis decision to publish or preparation of the manuscript

References1 World Health Organization Oral Health CountryArea

Profile Programme (CAPP) Geneva WHO 2006 Avail- able at ldquohttpwwwmahsecapprdquo Accessed June 12 2014

2 US Department of Health and Human Services Trend in Oral Health Status United States 1988-1994 And 1999-2004 Atlanta Ga Centers for Disease Control and Prevention National Center for Health Statistics 2007

3 Pitts NB Chestnutt IG Evans D White D Chadwick B Steele JG The dentinal caries experience of children in the United Kingdom 2003 Br Dent J 200625313-20

4 Speechley M Johnston DW Some evidence from Ontario Canada of a reversal in the dental caries decline Caries Res 199630423-7

5 Armfield JM Roberts-Thomson KF Slade GD Spencer AJ Dental Health Differences between Boys and Girls The Child Dental Health Survey Australia 2000 AIHW cat no DEN 131 Canberra Australia Australian Institute

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105

354 CARIES PREDICTION USING MICROBIOLOGICAL TESTS

PEDIATRIC DENTISTRY V 36 NO 4 JUL AUG 14

of Health and Welfare (Dental Statistics and Research Series No 31) 2004

6 Haugejorden O Birkeland JM Evidence for reversal of the caries decline among Norwegian children Int J Paediatr Dent 200212306-15

7 Gao XL Hsu CYS Loh T Koh D Hwarng HB Xu Y Dental caries prevalence and distribution among pre- schoolers in Singapore Community Dent Health 2009 2612-7

8 Stamm J Disney J Graves R Bohannan H Abernathy J The University of North Carolina Caries Risk Assess- ment Study I Rationale and content J Public Health Dent 198848225-32

9 Zero D Fontana M Lennon AM Clinical applications and outcomes of using indicators of risk in caries man- agement J Dent Educ 2001651126-32

10 Powell LV Caries prediction a review of the literature Community Dent Oral Epidemiol 199826361-71

11 Tellez M Gomez J Pretty I Ellwood R Ismail A Evi- dence on existing caries risk assessment systems are they predictive of future caries Community Dent Oral Epi- demiol 20134167-78

12 Leverett DH Proskin HM Featherstone JD et al Caries risk assessment in a longitudinal discrimination study J Dent Res 199372538-43

13 Gao XL Hsu CYS Xu Y Hwarng HB Loh T Koh D Building caries risk assessment models for children J Dent Res 201089637-43

14 Axelsson P Prediction of caries risk and risk profiles In Axelsson P eds Diagnosis and Risk Prediction of Dental Caries 2nd edition Chicago Quintessence Publishing Co 2000249-80

15 Thenisch NL Bachmann LM Imfeld T Leisebach Minder T Steurer J Are mutans streptococci detected in pre- school children a reliable predictive factor for dental caries risk A systematic review Caries Res 200640366-74

16 Barsamian-Wunsch P Park JH Watson MR Tinanoff N Minah GE Microbiological screening for cariogenic bacteria in children 9 to 36 months of age Pediatr Dent 200426231-9

17 Slayton RL Cooper ME Marazita ML Tuftelin mutans streptococci and dental caries susceptibility J Dent Res 200584711-4

18 Hegde PP Ashok Kumar BR Ankola VA Dental caries experience and salivary levels of Streptococcus mutans and lactobacilli in 13-15 years old children of Belgaum city Karnataka J Indian Soc Pedod Prev Dent 200523 23-6

19 Gudkina J Brinkmane A The impact of salivary mutans streptococci and sugar consumption on caries experience in 6-year-olds and 12-year-olds in Riga Stomatologija 20101256-9

20 Van Houte J Microbiological predictors of caries risk Adv Dent Res 1993787-96

21 Parisotto TM Steiner-Oliveira C Silva CM Rodrigues LK Nobre-dos-Santos M Early childhood caries and mutans streptococci a systematic review Oral Health Prev Dent 2010859-70

22 Silness J Loumle H Periodontal disease in pregnancy II Correlation between oral hygiene and periodontal condi- tion Acta Odontol Scand 196422121-35

23 WHO Oral Health Surveys Basic Methods 4th ed Geneva Switzerland WHO 1997

24 Bratthal D Dental caries In Johnson NW eds Markers of High- and Low-risk Groups and Individuals Cambridge UK Cambridge University Press 1991316-28

25 Lang C Boumlttner M Holz C et al Specific Lactobacillusmutans streptococcus co-aggregation J Dent Res 2010 89175-9

26 Bosch M Nart J Audivert S et al Isolation and charac- terization of probiotic strains for improving oral health Arch Oral Biol 201257539-49

27 Martiacutenez-Paboacuten MC Ramiacuterez-Puerta BS Escobar-Paucar GM Franco-Corteacutes AM Physicochemical salivary pro- perties Lactobacillus mutans streptococci counts and early childhood caries in preschool children of Colombia Acta Odontol Latinoam 201023249-56

28 Schroumlder U Edwardsson S Dietary habits gingival status and occurrence of Streptococcus mutans and lactobacilli as predictors of caries in 3-year-olds in Sweden Com- munity Dent Oral Epidemiol 198715320-4

29 Thibodeau EA OrsquoSullivan DM Tinanoff N Mutans streptococci and caries prevalence in preschool children Community Dent Oral Epidemiol 199321288-91

30 Twetman S Staringhl B Nederfors T Use of the strip mutans test in the assessment of caries risk in a group of preschool children Int J Paediatr Dent 19944245-50

31 Pienihaumlkkinen K Jokela J Clinical outcomes of risk-based caries prevention in preschool-aged children Community Dent Oral Epidemiol 200230143-50

32 Petersson GH Isberg PE Twetman S Caries risk assess- ment in school children using a reduced Cariogram model without saliva tests BMC Oral Health 201019105