Oral Bacterial Acid–Base Metabolism in Caries Screening: A Proof-Of-Concept Study

Abstract

The objective of this cross-sectional study was to clinically validate an array of biochemical tests as caries screening tools for oral acid–alkali generation. Adult subjects (n = 185; mean 33.6 ± 10.6 y) were examined clinically for dental caries using the International Caries Detection and Assessment System (ICDAS) criteria. Bitewing radiographs were used to confirm the interproximal surfaces of posterior teeth. For the purposes of this study, subjects were classified as “caries-active” if they had at least one untreated caries lesion with ICDAS 4 or higher. Pooled supragingival plaque and unstimulated saliva samples were collected and assayed for pH changes from sucrose and urea metabolism using colorimetric tests. The validity of each test to discriminate between “caries-inactive” and “caries-active” subjects was assessed and compared with a commercial bacteriological caries-screening test using roc regression and logistic regression models. The areas under the curve (AUCs) (95% CI) of the plaque-urea (PU, 0.59 (0.51 to 0.67)), plaque-urea-glucose (PUG: 0.59 (0.51 to 0.67)) and saliva-urea-glucose (SUG, 0.59 (0.51 to 0.67)) tests did not differ significantly from the bacteriological tests (CRT-mutans, 0.62 (0.54, 0.70); CRT-lactobacillus, 0.63 (0.56 to 0.71) (P > 0.05), but the plaque-glucose (PG), saliva-glucose (SG), saliva-urea (SU) and saliva-plaque-glucose (SPG) tests had significantly smaller AUCs (P < 0.05). The AUCs for PU, PUG, SUG, and the CRT-mutans tests were larger in subjects who had no existing dental restorations (PU, 0.90 (0.77 to 1.04); PUG, 0.90 (0.79 to 1.01); SUG, 0.89 (0.69 to 1.08); CRT-mutans, 0.90 (0.73 to 1.08)). The incorporation of the biochemical tests into a multidimensional bacteriological/psychosocial caries screening model significantly increased the diagnostic value (sensitivity and specificity, 160.6; AUC, 0.846). In conclusion, as a proof-of-concept, the results of this study indicate that measuring urea metabolism together with sugar metabolism by dental plaque and saliva may have a promising role in caries screening either independently or as part of a multidimensional biological test.

Knowledge Transfer Statement: The results of this study indicate that assessment of the oral acid/base balance may have a promising role in caries screening either independently, or as part of a multidimensional test.

Introduction

Reducing or eliminating disparities in the distribution and severity of dental caries requires targeted interventions (Zavras et al. 2000; Jokela and Pienihäkkinen 2003; Pienihakkinen and Jokela 2002). To implement such interventions, it is necessary to develop validated, low-cost, rapid screening tools that can identify individuals at risk. Currently, the best indicator of caries risk is previous caries experience; consequently, high-risk individuals can be identified only by experienced dental personnel and only after the disease has manifested into a relatively advanced stage (Berg 2006; Divaris 2016). An ideal caries-screening instrument should be founded on biological caries determinants that can be detected at the pre-clinical stage and that can be assessed without a clinical examination.

Commercial biological caries activity tests focus on the capacity of dental plaque bacteria to generate acids from sugars, primarily by assessing the numbers of highly acidogenic species (Baehni and Guggenheim 1996; Baca et al. 2011). However, recent meta-genomic studies suggest that various microbial consortia may be collectively responsible for caries initiation and progression in different individuals (Takahashi 2011; Simón-Soro and Mira 2015). These caries-inducing consortia may vary among individuals and their metabolic activities appear to be highly regulated in response to various environmental conditions (Benítez-Páez et al. 2014). Therefore, from a diagnostic view point, assessing the overall metabolic activity of microbial communities in dental plaque would likely be more informative in predicting caries activity and risk than assessing the microbial composition (Simón-Soro and Mira 2015). Consistent with this concept, this report illustrates a new approach for caries screening and, potentially, risk assessment founded on the Caries Risk Pyramid (CRP), a hierarchical model describing the biology of caries development (Morou-Bermudez, Billings et al. 2011).

The CRP model postulates that acid production from sugars and alkali production from endogenous sources of bacteria in dental plaque and saliva summarize the combined effect of all other biological and psychosocial caries risk factors (Morou-Bermudez, Billings et al. 2011). A simple, rapid method to measure simultaneously these opposing activities could potentially enhance the accuracy of a screening instrument for determining caries activity and risk. Among the various sources of alkali in the oral cavity, urea is the most abundant (Kleinberg 2002; Takahashi 2015). Recent clinical studies have shown that reduced levels of urease activity in plaque are associated with increased caries levels in adults (Shu et al. 2007; Nascimento et al. 2009) and increased risk for caries development in children (Morou-Bermudez et al. 2011b). Furthermore, urease activity appears to be linked to sugar consumption and to levels of mutans streptococci (Morou-Bermudez et al. 2011a). These observations suggest that alkali generation from urea can be a potentially important new risk factor not previously considered in caries screening or risk assessment.

To rapidly and concomitantly assess the balance between oral acid and base, we developed an array of simple colorimetric tests that measure alkali generation from urea and other endogenous sources by bacteria in dental plaque and saliva, as well as acid production from sugars, separately or simultaneously. The short-term objective of this study was to determine the validity and reliability of these tests to identify individuals with untreated caries. As a proof-of-concept, this study focused on adults with relatively advanced lesions (International Caries Detection and Assessment System [ICDAS] 4 and higher). The long-term objective of our research is to develop a unique, multidimensional biological test to predict dental caries in the earliest—ideally, pre-clinical—stage in all populations, especially very young children at high risk for severe early childhood caries.

Materials and Methods

The methodology of this study conforms to the STROBE Guidelines.

Study Design and Population

This was a cross-sectional study on a convenience sample of 185 adult subjects (21 to 61 y old). The sample size was estimated using preliminary data from a previous study to detect significant differences in plaque urease activity between clinically caries-free subjects and subjects with at least one untreated caries lesion, with a level of significance at 0.05 and 80% power for a range of ratios of “caries active” vs. “caries inactive” between 0.25 to 4.0 (Morou-Bermudez et al. 2010a; 2010b). Participants were recruited using flyers within the University of Puerto Rico Medical Sciences Campus (UPR-MSC) between January 2012 and September 2012. The inclusion requirement was the presence of at least one anterior and one posterior tooth in each quadrant of the mouth. The exclusion criteria included medical conditions requiring pre-medication for routine dental treatment, bleeding of the gums, treatment with antibiotics within the previous 2 months, medications affecting salivary flow, hyposalivation, immunosuppression, orthodontic or prosthetic appliances, smoking, hormonal disturbances, pregnancy, and breastfeeding. Eligible participants provided written informed consent, which was approved by the IRB of the UPR-MSC (Protocol # A0060111).

Clinical Procedures

Participants were instructed to refrain from oral hygiene procedures for 24 h before the appointment and to only consume water from the night before. Participants answered a brief sociodemographic questionnaire and provided a 24-h diet recall, which was used to quantify sugar consumption (Toro et al. 2010). The participants rinsed their mouth with water for 30 s immediately before sample collection. About 3 ml of whole unstimulated saliva was collected. Subsequently, the teeth of each quadrant were rinsed for 5 s with water spray, isolated with cotton rolls, and dried for 5 s with air spray. Supragingival plaque was collected from the buccal surfaces of one molar and one incisor tooth in each quadrant and pooled into a pre-weighed, micro-centrifuge tube. All sample collections took place between 7 am and 10 am. Saliva and plaque samples were kept on ice and transferred to the adjacent laboratory within 1 h of collection. The dental assistant brushed the participant’s teeth before the dental examination.

Dental exams were performed using the ICDAS criteria (Ismail et al. 2007). One calibrated examiner performed all exams. A training and calibration exercise was conducted at the start of the study on 30 subjects for a total of 4,703 tooth surfaces. The inter-rater reliability (% Agreement and weighted kappa ± SE) as calculated on the last 5 patients of the training/calibration exercise (785 surfaces) was 97.94% and 0.72 ± 0.03 (P < 0.001). The intra-rater reliability was evaluated periodically during the study on 10% of the subjects; intra-rater kappa was 0.62 (98.39% agreement). Four digital bitewing radiographs were taken in the radiology clinic of the UPR School of Dental Medicine, unless the participant could provide recent (within 1 to 2 y, depending on the clinical caries status of the participant) bitewings of good diagnostic value. The bitewings were used for diagnostic as well as research purposes. All X-rays were evaluated by the same examiner using the ICDAS criteria for X-rays (ICDAS Foundation, 2015). When a disagreement existed between the clinical exam and the X-rays on an interproximal surface, the X-ray score predominated (Llena-Puy and Forner 2005). For all other dental surfaces, the clinical score predominated over the X-ray score (Diniz et al. 2011).

Subjects were classified as “caries active” (CA) if they had at least one untreated caries lesion with dentinal involvement (clinical ICDAS score ≥4 and/or X-ray ICDAS score ≥3) (ICDAS Foundation, 2015); this definition included recurrent lesions under existing or missing restorations. All other subjects were classified as “caries inactive” (CI), including those with ICDAS scores 1 through 3 (lesions restricted to the enamel), those with restorations with no clinical or radiographic evidence of recurrent decay, and teeth extracted due to caries. Lesions with ICDAS 1 through 3 are not necessarily “inactive” but they tend to progress at much lower rates as compared with lesions scoring ICDAS 4, and they do not usually require surgical interventions (Ferreira Zandona et al. 2012; Pitts and Ekstrand 2013).

To assess the reliability of the tests, a second plaque and saliva sample was collected at 1 to 4 weeks after the first visit using the same methodology.

Biochemical and Microbiological Procedures

Acid and alkali production in dental plaque and saliva was measured with custom-made colorimetric tests. PU (Plaque-Urea) and SU (Saliva-Urea) tests measured pH changes from the metabolism of 10 mM urea in plaque and saliva, respectively; PG (Plaque-Glucose) and SG (Saliva-Glucose) tests measured pH changes from the metabolism of 10 mM glucose in plaque (PG) or saliva (SG); PUG (Plaque-Urea-Glucose) and SUG (Saliva-Urea-Glucose) tests measured pH changes from the simultaneous metabolism of 10 mM urea and 10 mM glucose in plaque or saliva. To consider all other possible alkali-generating substrates in saliva, we used the SPG (Saliva-Plaque-Glucose) test. This test was similar to the PG test but it also included clarified saliva from the same patient. Control solutions with no sample and no substrate were included to determine background pH changes.

Tests were performed immediately after collecting plaque and saliva samples. The amount of plaque used for PU, PG, PUG, and SPG tests was standardized by wet weight (0.1 mg); saliva for the SU, SG, and SUG tests was standardized by volume (50 µl). Plates were covered with sealing tape and incubated at 37°C. Test results were recorded after 1 h, 2 h, 3 h, and overnight. The score for each test at each time point corresponded to the pH reading of the test minus the pH change of the control solution. The pH was determined in 0.25-unit increments using a colorimetric pH scale (Phenol Red). All tests were performed and scored by one calibrated technician, who was blind to the caries status of the subjects. The correlation (Spearman Rho) between the visual color scale measurements and a calibrated pH meter (SympHony B20, VWR, Radnor PA) was 0.92 (P < 0.001), as determined by a pilot calibration experiment.

A commercial bacteriological test for salivary mutans streptococci and lactobacilli (Caries Risk Test [CRT]-bacteria, Ivoclar Vivadent, Liechtenstein) was included for comparison against the biochemical tests. These caries risk tests were performed with 1 ml saliva each, according to the manufacturer’s instructions. The results from these tests were read after 48 and 96 h, and data are expressed as a categorical variable with scores between 1 (≤10⁵ CFU/ml) and 4 (>10⁵ CFU/ml), according to the manufacturer’s instructions. Total salivary bacterial load (CFU/ml) was assessed by plating on duplicate brain heart infusion (BHI) plates (Fluka-Sigma-Aldrich) after 48 h anaerobic incubation, followed by 24 h aerobic incubation.

Analytic Procedures

Summary statistics were used to describe the study group. Chi-squared distribution and Mann–Whitney tests were used to compare CA and CI groups according to demographic and clinical characteristics. Nonparametric receiver operating characteristics (ROCs) were initially calculated for each test at each time point. The 3-h results for the biochemical tests and the 96-h scores for the bacteriological tests had the best score distribution and the best ROCs as compared with the other time-points, and were therefore selected for further analysis. The validity of each test was assessed using sensitivity, specificity, positive and negative predictive values (PPV and NPV), and roc regression adjusted for age and gender. The scores of each biochemical test (which corresponded to the adjusted pH values at each time point) were inverted for this analysis, as higher pH values indicate lower risk. To determine which combination of tests best explains the caries status, a multivariate analysis was performed using logistic regression models. The diagnostic value of these multivariate models was assessed with the Stata commands “lroc” and “estat classification”. The likelihood ratio (LR) test was used to assess interactions and to compare different multivariate logistic regression models for the classification of CA and CI subjects (Hosmer and Lemeshow 2000). The reliability of each test between the 2 visits was evaluated using weighted kappa statistics. Data were analyzed using STATA version 12.0 software.

Results

The study group consisted of 81 males and 104 females aged between 21 and 61 y (mean 33.6 ± 10.6 y). Of these, 81 subjects (43.78%) were considered “caries active” (CA) according to the study criteria, while 104 subjects (56.22%) were “caries inactive” (CI) (Table 1). There were no significant age or gender differences between the 2 groups (P > 0.05). The CA group had significantly fewer subjects with college-level education (P = 0.005) and fewer subjects with private dental insurance (P = 0.035). CA subjects had significantly higher numbers of cavitated lesions restricted to the enamel (ICDAS score 3) as compared with CI subjects (P = 0.025) but there was no significant difference in the numbers of noncavitated enamel lesions (ICDAS scores 1 and 2) or the numbers of existing restorations between the CA and CI subjects. CA subjects had significantly higher salivary loads of mutans streptococci and lactobacilli compared with CI subjects (P < 0.01). No significant differences were observed between the CA and CI subjects with respect to the amount of plaque, sugar consumption, and total bacterial load in the saliva (P > 0.05).

Table 1.

Psychosocial, Biological, and Clinical Characteristics of the Study Group by Caries Status.

	Caries Inactive (n = 104)	Caries Active (n = 81)
	% or Mean ± SD (Median)		P Value
Gender	56.2% males	43.8% males	>0.05a
Age	33.6 ± 10.8 (30) y	33.6±11 (31) y	>0.05b
Education	62.8% college	37.2% college	0.005a
Dental insurance	63.4% private	36.6% private	0.035a
Frequency of dental visits	78% every 6 months	65% every 6 months	0.060a
Frequency of brushing	90% more than twice a day	83% more than twice a day	>0.05a
Sugar consumption	3.9 ± 2.4 (3.75)	3.8 ± 2.1 (4)	>0.05b
Amount of plaque	8.9 ± 4.2 (8) mg	9.7 ± 3.9 (9.5) mg	0.083b
Saliva CFU	90.1 ± 72.9 (74)	101.3 ± 102.5 (87)	>0.05b
Salivary mutans (CRT score)	1.75 ± 0.86 (2)	2.14 ± 1.00 (2)	0.008b
Salivary lactobacilli (CRT score)	1.53 ± 0.62 (1)	1.91 ± 0.90 (2)	0.005b
Enamel, noncavitated (ICDAS 1 and 2)	6.5 ± 5.1 (5)	7.5 ± 6.8 (5)	>0.05b
Enamel, cavitated (ICDAS 3)	1.1 ± 1.6 (0)	1.4 ± 1.5 (1)	0.025b
Dentine (ICDAS 4 to 6)	0	2.6 ± 2.6 (2)
Total restorations	14.4 ± 11 (13)	14.2 ± 12.4 (13)	>0.05b
DFS	9.96 ± 8.64 (9)	13.4 ± 11.05 (12)	0.027b

CRT, CRT-bacteria tests (Caries Risk Tests), Ivoclar Vivadent, Liechtenstein; CFU, colony forming units; DFS, decayed and filled surface; ICDAS, International Caries Detection and Assessment System criteria (Ismail et al. 2007).

^aPearson χ2.

^bMann-Whitney test.

In the plaque samples, urea metabolism produced on average a 0.7-unit increase in the pH within the first 3 h (Fig. 1B), whereas the metabolism of an equal molar amount of glucose had a very small effect on the pH within the first 3 h (on average 0.05 pH units) (Fig. 1D). Consequently, when both substrates were available to the plaque samples at the same time, there was an initial increase in the pH (about 0.5 pH units in the first 3 h) and the final pH after overnight incubation returned to neutral levels (Fig. 1F). In the saliva samples, urea metabolism produced an average 0.32-unit increase in the pH in the first 3 h (Fig. 1A), whereas the metabolism of an equal concentration of glucose produced an average 0.5-unit drop in the pH during the same period (Fig. 1C). When both substrates were available simultaneously to the saliva samples, no change was observed in the pH during the first 3 h, and only a very small pH drop (mean, 0.17 units) was observed after the overnight incubation (Fig. 1E). The pattern of the SPG test was closer to that of the SG test compared with the PG test (Fig. 1G). The trend of the PUG test over time was significantly different between CI and CA subjects (P = 0.015). No differences were observed in the trends of the PU, PG, SU, SG or SUG tests between the 2 groups.

Figure 1.

Observed pattern of pH change in the biochemical tests over time in the caries active and caries inactive groups.

The sensitivity of the biochemical tests ranged between 14.8% (PG) and 89.8% (SUG), and the specificity between 18.9% (SPG) and 84.6% (PG) (Table 2). The bacteriological tests had a sensitivity of 50.0% to 51.9% and specificity of about 57.0%. The PUG test (127.5%) had the highest sum of sensitivity plus specificity whereas the SG test (86.4%) was the lowest. The PPVs were between 38.6% (SPG) and 53.9% (SU), and NPVs ranged between 44.8% (PU) and 85.0% (PUG).

Table 2.

Validity and Reliability of Biochemical and Bacteriological Tests in this Study (n = 185).

Test	Sensitivity (95% CI)	Specificity (95% CI)	Positive Predictive Value (95% CI)	Negative Predictive Value (95% CI)	Sensitivity + Specificity	ROC (95% CI), Adj. Age, Gender P Value Compared with CRT-LB		Kappa (% Agreement between Visits)
PUa	34.1(25.4, 42.8)	68.6(60.1, 77.1)	40.5(31.5, 49.6)	62.3(53.4, 71.2)	102.7	0.58(0.50, 0.66)	>0.05	0.40(88.4)
PGa	14.8(9.7, 19.9)	84.6(79.4, 89.8)	42.9(35.7, 50.0)	56.1(48.9, 63.2)	99.4	0.50(0.41, 0.58)	0.023	0.26(94.6)
PUGa	87.0(78.8, 95.1)	40.5(28.5, 52.4)	44.4(32.4, 56.5)	85.0(76.3, 93.7)	127.5	0.59(0.51, 0.67)	>0.05	0.44(92.0)
SPGa	75.0(66.0, 84.0)	18.9(10.7, 27.0)	38.6(28.5, 48.7)	52.6(42.3, 63.0)	93.9	0.48(0.40, 0.56)	0.009	0.23(93.5)
SU	57.1(47.8, 66.5)	58.6(49.3, 68.0)	53.9(44.4, 63.3)	61.8(52.6, 71.0)	115.7	0.51(0.43, 0.60)	0.035	0.44(95.5)
SUa	68.8(58.5, 79.0)	21.7(12.6, 30.9)	37.9(27.2, 48.7)	50.0(38.9, 61.1)	90.5	0.49(0.40, 0.57)	0.030
SGa	56.8(46.0, 67.6)	29.6(19.6, 39.5)	40.4(29.7, 51.1)	44.8(34.0, 55.7)	86.4	0.49(0.41, 0.57)	0.005	0.46(96.1)
SUGa	89.8(84.9, 94.8)	27.1(19.8, 34.3)	46.1(38.0, 54.2)	79.3(73.0, 85.9)	116.9	0.60(0.53, 0.68)	>0.05	0.40(94.9)
CRT-MS	51.9(43.6, 60.1)	57.0(48.8, 65.2)	43.1(34.9, 51.3)	65.3(57.5, 73.2)	108.9	0.61(0.53, 0.69)	>0.05	0.65(92.3)
CRT-LB	50.0(42.2, 57.8)	57.3(49.6, 65.0)	43.1(35.3, 50.8)	64.0(56.5, 71.4)	107.3	0.63(0.55, 0.71)		0.65(95.3)

CRT-MS, CRT test for mutans streptococci; CRT-LB, CRT test for lactobacillus; PG, Plaque-Glucose; SG, Saliva-Glucose; PU, Plaque-Urea; SU, Saliva-Urea; PUG, Plaque-Urea-Glucose; SUG, Saliva-Urea-Glucose; SPG, Saliva-Plaque-Glucose; ROC, receiver operating characteristics.

^aTest scores were inverted so that higher scores could correspond to higher risk. Scores with low frequency were regrouped (collapsed) to obtain equal groups. The scores for SG were divided in quartiles for the same reason.

The largest area under the curve (AUC) was for the CRT-lactobacillus test (0.63; 95% CI = 0.55 to 0.71). The AUCs of the CRT-mutans, PU, PUG and SUG tests were between 0.58 and 0.61, and they did not differ significantly from the CRT-lactobacillus test (P > 0.05; Fig. 2A, C; Table 2); the remaining tests (PG, SU, SPG, and SG) were significantly lower compared with the CRT-lactobacillus test (P < 0.05). The ROC of the SU test was inversed relative to the ROCs of the other biochemical tests.

Figure 2.

Age- and gender-adjusted ROCs of the biochemical and bacteriological tests for identifying subjects with at least one untreated caries lesion with dentin involvement in the study group (A and C, n = 185), and in the subjects who do not have dental restorations (B and D, n = 17).

The AUCs for the PU, PUG, SUG, and the CRT-mutans tests were much higher in 17 subjects, who did not have existing dental restorations [PU: 0.90 (0.77, 1.04); PUG: 0.90 (0.79, 1.01; SUG: 0.89 (0.69, 1.08); CRT-mutans: 0.90 (0.73, 1.08); Fig. 2B, D]. These subjects (10 CI, 7 CA) did not differ significantly from the rest of the group with respect to age, gender, sugar consumption, amount of plaque, total bacterial load in saliva, number of mutans streptococci in saliva, or number of untreated caries (P > 0.05). The ROCs of the PG and CRT-lactobacillus test were significantly smaller than the CRT-mutans test (P < 0.05) in these subjects.

Test scores that were significantly associated with caries status were the following: PUG scores ≥8.0 (adjOR: 0.20, 95% CI = 0.05 to 0.76); SUG scores ≥7.25 (adjOR 0.23, 95% CI = 0.08 to 0.69), SPG scores =6.5 (adjOR: 3.09, 95% CI = 1.23 to 7.59), CRT-mutans score >2 (adjOR: 2.86, 95% CI = 1.32 to 6.20), and CRT-lactobacillus scores >2 (adjOR: 4.81, 95% CI = 1.82 to 12.75). Based on these data, each subject was assigned one “Biochemical Risk” point for each of the following situations: PUG <8.0, SUG <7.25, SPG =6.5 and one “Bacteriological Risk” point for each of the bacteriological tests (CRT-mutans and CRT-lactobacillus) with a score ≥3. The sum of biochemical and biological risk points for each participant was defined as “Biological Risk”. The “Biochemical Risk” model (M1) had significantly better diagnostic values than the “Bacteriological Risk” (M2) (LR test P < 0.001) (Table 3). The “Biological Risk” (M3) had significantly better diagnostic value than the “Biochemical” (M1) and the “Bacteriological” (M2) risks (P < 0.001) but it was not significantly better than a model composed of simple psychosocial factors, including age, gender, education, insurance and frequency of dental visits (M4: P > 0.05). When the “Biological Risk” was combined with the psychosocial model, the resulting model (M5) had an ROC of 0.847, a combined sensitivity/specificity of 160.2, and it would correctly classify about 82% of the subjects according to their true caries status. This model (M5) was significantly better than the psychosocial (M4) (LR test, P < 0.001) and the model only factoring in Biological Risk (M3) (LR test, P = 0.011). The addition of any further tests, such as PU, SU, SG, and PG tests, or other biological and psychosocial variables, did not significantly improve the diagnostic values of the model. In a secondary analysis, the diagnostic value of the models was assessed using ICDAS 3 as the caries threshold (Table 3). At this threshold, 71.35% of the subjects were classified as CA. When the caries threshold was reduced, the sensitivity of all tests increased but the specificity decreased, reducing the overall diagnostic value of the models. The biochemical tests had higher specificity and overall diagnostic values at the ICDAS 3 threshold as compared with the bacteriological tests.

Table 3.

Screening Values of Multivariate Models for Caries Activity at the ICDAS 4 and ICDAS 3 Thresholds.

Model	Variables	Caries Threshold	Se	Sp	PPV	NPV	CC	Se+Sp	AUC
M1: Biochemical	“Biochemical Risk”a	ICDAS 4	37.0	81.7	61.2	62.5	62.16	118.7	0.660
		ICDAS 3	90.2	32.1	76.8	56.7	73.51	122.3	0.682
M2: Bacteriological	“Bacteriological Risk”b	ICDAS 4	38.3	76.9	56.4	61.5	60.00	115.2	0.576
		ICDAS 3	100.0	0.0	71.4	NCc	71.35	100.0	0.523
M3: Biological	M1+M2	ICDAS 4	32.1	99.0	96.3	65.2	69.73	131.1	0.745
		ICDAS 3	92.4	26.4	75.8	58.3	73.61	118.8	0.673
M4: Psychosocial	Age, gender, education, insurance, frequency of dental visits	ICDAS 4	50.0	85.7	72.0	70.0	70.59	135.7	0.695
		ICDAS 3	95.8	19.6	73.6	66.7	72.94	115.4	0.680
M5: Comprehensive	M3+M4	ICDAS 4	70.83	89.80	83.61	80.73	81.76	160.6	0.846
		ICDAS 3	90.8	41.2	78.3	65.6	75.88	132.0	0.753

Se, sensitivity; Sp, specificity; PPV, positive predictive values; NPV, negative predictive values; CC, correctly classified; AUC, area under the curve.

^aThis variable is defined as 1 point for each of the following results: PUG, <8.0; SUG <7.25; SPG, =6.5 (0 to 3).

^bThis variable is defined as 1 point for each score 3 and above (0 to 2).

^cCannot be calculated.

Likelihood ratio test to compare different models (at the ICDAS 4 threshold): M1 vs. M2, P < 0.001; M1 vs. M3, P < 0.001; M2 vs. M3, P < 0.001; M3 vs. M4, P > 0.05; M3 vs. M5, P = 0.035; M4 vs. M5, P < 0.001.

The reliability of the PU, PUG, SU, SUG and SG tests between the two visits was average, with weighed kappa values ranging from 0.40 (PU, SUG) to 0.46 (SG) (Table 2). The reliability of the PG and SPG test was poor (kappa: 0.26, 0.23, respectively). The results of the bacteriological tests were more consistent between the 2 visits (kappa: 0.65).

Discussion

This study presents a novel approach for caries screening based on the emerging evidence that effective monitoring of the acid–base physiology of dental plaques must be considered in caries risk assessment (Burne and Marquis 2000; Nascimento et al. 2009; Morou-Bermudez et al. 2011b). Our caries-screening method incorporates a new risk factor not previously considered in caries screening or risk assessment, specifically, the generation of alkali from urea and other endogenous nitrogen sources by bacteria in dental plaques and saliva. According to the Caries Risk Pyramid Model, this activity should be evaluated together with the process of acid production in order to assess caries activity and caries risk more effectively (Morou-Bermudez, Billings et al. 2011).

The rationale for selecting urea as the alkali-generating substrate is because urea is more abundant in the oral cavity than other nitrogenous substrates (Kopstein and Wong 1977; Takahashi 2015). Furthermore, oral ureolysis is the only oral alkali source that has been examined longitudinally with respect to caries development (Morou-Bermudez et al. 2011b). Overall, the biochemical findings of the study were in agreement with observations from previous studies (Kleinberg et al. 1973; Shu et al. 2007; Nascimento et al. 2009). The diagnostic value of the individual PU, PUG, and SUG tests equaled that of the bacteriological tests; however, our tests can be conducted much faster than the bacteriological tests (in 3 h vs. 48 h), and this offers a clinical advantage. The incubation time of the biochemical tests can be further reduced by increasing the concentrations of substrates or the sample and by using more sensitive pH measuring technologies in place of the colorimetric method (Demuth et al. 2016). The use of a more streamlined technology could also improve the low repeatability observed between the 2 visits. The variability in the results between the 2 visits could also be explained by the fact that the expression of the genes involved in sugar and urea metabolism in oral bacteria is highly regulated by carbohydrate availability, nitrogen availability, and pH (Morou-Bermudez and Burne 2000; Burne and Marquis 2000; Lemos et al. 2005; Liy et al. 2008). To address this issue, we used a standardized sample collection protocol that was performed under fasting conditions; however, it is possible that the participants followed the fasting requirement only for the first visit and not the second one. Our study design controlled for other possible sources of variability, such as the use of antibiotics, smoking, immunological and hormonal disturbances that can significant affect the composition and metabolic activity of the dental biofilm; therefore, we do not know the impact of these variables on the test accuracy. Once these limitations have been addressed, this simple technology has the potential to be adapted into a fast, reliable, low-cost, chair-side instrument for professional or personal use.

The diagnostic values of the biochemical and the bacteriological tests observed in this study were lower than those reported by others (Edelstein et al, 1989; Leverett et al. 1993; Thibodeau et al. 1993; Baca et al. 2001; Nishimura et al. 2008; Yoon et al. 2013); this is not surprising, given that our study group consisted of adults with a range of previous caries and possible dental restorations. Biological tests are generally reported to have poor diagnostic value in older children (older than 6 y) and in adults (Baehni and Guggenheim 1996; Zero et al. 2001; Pinelli et al. 2001). It has also been proposed that the diagnostic value of biological caries risk tests depends largely on the caries experience of the test population (Zero et al. 2001). In our study, however, it appeared that both the biochemical and bacteriological tests are more accurate in subjects who did not have existing restorations, even if these subjects had lost teeth in the past due to caries. Although this was a very small group, the ROCs obtained were significant, as evidenced by the 95% CIs. Our results were not different when secondary caries was excluded from the analysis (data not shown). Although based on a small sample, these observations suggest that caries screening methods founded on biological determinants may be more accurate in subjects who have no other “iatrogenic” risk factors.

The PUG and SUG tests, which contain both urea and glucose, had the highest sum of sensitivity and specificity compared with the PU, SU and CRT-tests, suggesting that the simultaneous measurement of acid and alkali production in plaques may be a more accurate indicator of caries activity than measuring each of these activities separately. The PUG, SUG, and SPG tests had high sensitivity compared with CRT tests (75% to 90% vs. 50%), but their specificity was low. The low specificity reflects the fact that these tests assess the overall metabolic activity of plaque, and not specific bacteria. However, at an earlier stage (ICDAS 3), the biochemical tests showed higher specificity and better overall diagnostic value compared with the bacteriological tests in detecting caries lesions. By combining the biochemical and bacteriological tests, we obtained an increased specificity (99%), as expected, but the sensitivity was reduced. The further incorporation of a few simple psychosocial factors (age, gender, education, insurance and frequency of dental visits) produced a model with a combined sensitivity plus specificity of at least 160%, which is considered acceptable for clinical use (Stamm et al. 1988; Zero et al. 2001), and a significantly better overall diagnostic value. The sensitivity of the model was 71%, which means that about one-third of the “caries-active” subjects remained undetected. Based on the available literature and the observations from this study, it is possible that the sensitivity of this model may be higher in small children, in populations with low caries experience, and in subjects without existing restorations (Zero et al. 2001).

As a proof-of-concept, this study validated the utility of various biochemical tests for oral acid–alkali generation as caries screening tools in adults with previous caries experience. Our findings demonstrate that urea and urea-sucrose tests equaled the commercially available bacteriological tests in our study population, and that the incorporation of these tests into a multidimensional, biological, diagnostic model significantly improved its diagnostic value. Further studies are planned to assess and improve the accuracy, reliability, effectiveness and efficiency of these tests to detect caries lesions at earlier stages and in higher-risk populations, including children with no prior exposure to oral health care system.

Author Contributions

E. Morou-Bermudez, contributed to conception, design, data analysis, and interpretation, drafted the manuscript; M.A. Loza-Herrero and V. Garcia-Rivas, contributed to data acquisition, critically revised the manuscript; E. Suarez-Perez and R.J. Billings, contributed to data analysis and interpretation, critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.