Clinical validation of a novel bioluminescence imaging technology for aiding the assessment of carious lesion activity status

Abstract Objectives Clinical validation of a bioluminescence imaging system (Cis) as measured by the level of agreement between clinician visual and tactile assessment of carious lesion presence and activity and the presence/absence of elevated luminescence on a tooth surface determined from intraoral image mapping. Materials and Methods This was a regulatory clinical study designed in consultation with the FDA. The design was a prospective, five‐investigator, nonrandomized, post‐approval, clinical study utilizing the Cis to provide images of elevated calcium ion concentration (indicative of active demineralization) on tooth surfaces via use of a photoprotein. Imaged teeth were identified as “sound” or having “active lesions.” Images were scored independently for luminescence. Results A total of 110 participants aged 7–74 years were imaged. Of the 90 teeth assessed as “sound,” 88 were deemed to show no luminescence by the reviewing investigator, a negative percentage agreement of 97.8% (significantly >50% agreement [p < .0001]; one‐sided 97.5% confidence interval [CI]: 0.9220). Of the 86 teeth initially assessed as having an “active lesion,” 78 were deemed to show luminescence by the reviewing investigator, a positive percentage agreement of 90.7% (significantly >50% agreement [p < .0001]; 97.5% CI: 0.8249). There were no patient‐related adverse events. Conclusions Results show, with a high level of agreement, that Cis can differentiate tooth surfaces clinically identified as involving active enamel lesions (ICDAS code 2/3), from sound sites (biochemically equivalent to inactive lesions) and that the system is safe for clinical use.


| INTRODUCTION
Dental caries (tooth decay) prevalence varies worldwide from 15 to 84% and while currently declining, it is still a significant clinical and public health problem (Frenken et al., 2017;Pitts et al., 2017). The development of caries lesions involves a net mineral loss from dental hard tissue, mediated by acid diffusing from bacterial dental plaque on tooth surfaces. This localized demineralization may, or may not, lead to progressive loss of tooth structure and associated pain and morbidity (Pitts et al., 2017;Selwitz et al., 2007).
Detecting, assessing, diagnosing, and treating carious lesions is a core activity in dentistry (Fejerskov et al., 2008;Paris et al., 2013;Pitts, 2009). The main detection and diagnostic aids have long been visual assessment of subtle differences in optical and physical characteristics of the lesion surface, including the use of a blunt ball-ended probe, together with radiographs (Neuhaus & Lussi, 2018). Assessment of a lesion's "true" activity status in a single visit remains challenging and determination of which "white spot" lesions will progress toward cavitation can currently only be validated prospectively by serially monitoring, over a specific time period, which lesions eventually undergo cavitation (Ekstrand & Martignon, 2013;Neuhaus & Lussi, 2018;Nyvad et al., 1999).
Several technologies can aid caries lesion detection; however, determination of the activity status of an individual lesion (one with ongoing net demineralization) is also required to fully assess treatment needs and success. Current methods of activity assessment are problematic and involve the dental professional's subjective judgment (Neuhaus & Lussi, 2018). There is therefore a need to develop a technique to aid in identifying the activity status of caries lesions so as to optimize the use of noninvasive preventive therapies, as compared to operative (surgical removal of dental tissue) interventions that tend to be more expensive and can have long-term negative clinical consequences, such as entry into the "restoration-rerestoration cycle" Schwendicke et al., 2018) as well as requiring the production of aerosols in a post COVID-19 pandemic world.
Whilst a number of different technologies, such a quantitative light fluorescence (QLF) and polarization-sensitive optical coherence tomography (PS-OCT), have shown indications that they may have some limited value in monitoring lesion activity, these have mostly been in a laboratory setting, that is, in vitro, and none is yet in general use in vivo (Fried, 2019;Kim, 2019).
A novel bioluminescence technology, the CALCIVIS Imaging System, has been developed to aid in the assessment of caries lesion activity. An in vitro study on occlusal sites of extracted teeth, using a pH dye for validation of activity, demonstrated: (a) a sensitivity of 92.5%; (b) a specificity of 90%, and (c) a diagnostic accuracy for activity assessment of caries lesions which was not significantly different from the value obtained using the ICDAS activity assessment method (Jablonski-Momeni et al., 2018), which was the validated activity assessment method used in the current article-vide infra. The technology underlying the CALCIVIS imaging system was developed to help address this unmet need. It combines a sensitive intraoral imaging device and a photoprotein solution that produces luminescence, or light emission (low level in the visible spectrum), in the presence of calcium ions in solution, as they are released from actively demineralizing regions of a tooth surface, that is, in the enlarged openings of the enamel pores of the surface zone of an enamel lesion. The biochemical reaction is provided below (Longbottom & Vernon, 2019).
3 × Ca 2 + + photoprotein = Apo− protein + Cofactor + CO 2 + blue light in region of 480nm ð Þ : The light emission results from a chemical reaction between the photoprotein and free calcium ions, in contrast to fluorescence-based technologies, which require an excitatory light source (Neuhaus & Lussi, 2018

| Context for studies of carious lesion activity
One time-point in vivo assessment of the activity status of a lesion using a visual-tactile method is generally recognized as being subjective and is limited by the examiner's visual acuity and tactile sensitivity. One of the visual-tactile methods with supporting peer-reviewed publications, the Nyvad criteria (Nyvad & Baelum, 2018;Nyvad et al., 1999Nyvad et al., , 2003, for example, when used by highly trained and calibrated examiners, was shown to have a reliability of 68.7% when used to designate noncavitated lesions as "active" and 72.5% to designate noncavitated inactive lesions-meaning that there was between a 1-in-3 and 1-in-4 chance that a lesion "active" or "inactive" designation would "transition" to the other designation when the same lesion was examined subsequently by the same examiner. Therefore, attempting to measure the performance of any technology designed to aid this diagnostic task of lesion activity assessment by using one of the current visual-tactile methods as a reference point (or "Gold standard") presents a substantial methodological challenge. As noted above, the current visual/tactile methods of lesion activity assessment, such as the Nyvad method (or the ICDAS activity assessment technique (Ekstrand et al., 2009)), fall considerably short of acceptable accuracy, hence, any study which uses either of these methods as a simple "gold standard" one time-point validation method faces producing unknowable levels of false positive (FP) and false negative (FN) results for any new technique under investigation. Thus, using such a simple comparative methodology, any new technology can only ever be as good as or worse than, in terms of accuracy, the supposed "gold standard." Hence, such a methodology is an exercise in futility-why carry out such a study when the test technique cannot possibly be shown to improve upon the knowingly "invalid"/"limited" reference standard?
The crux of the problem of determining as accurately as possible the "true" gold standard "activity status" of a lesion in vivo lies in a methodological paradox. If a new test technology operates at a molecular/ionic level, independent of the macro-morphology of the enamel, as is the case with the Calcivis technology, using as a reference/validation standard a method which operates at a macromorphological level (i.e., visual-tactile) means that the new technology is operating at resolution which is several orders of magnitude greater sensitivity than the reference method-the enamel pores of an active lesion are of the order of hundreds of nanometers to a micron in diameter (Haikel et al., 1983;Thylstrup et al., 1983) well beyond the limit of the human visual acuity used in the visual-tactile method.
All this demonstrates that there is a major methodological challenge in attempting to validly test the performance of the Calcivis imaging system in aiding the identification of active lesions compared with nonactive surfaces.
However, for an initial assessment, it is possible to reduce the probability of these unknown "validator errors" risks occurring by using, as comparative test samples, tooth surfaces which are known, from large numbers of epidemiological studies, to have the maximum difference in probability of exhibiting active caries lesions. Such sites in permanent teeth are: for least probability sites, the buccal surfaces of canines (or incisors) and for highest probability sites, the occlusal (pit and fissure) surfaces of molars (and premolars) (Batchelor & Sheiham, 2004;Massler et al., 1954;Poulsen & Horowitz, 1974).
Thus, designation by a clinician (using visual/tactile criteria) of a site on the buccal surface of a canine (or incisor) as "sound" has the highest probability of being correct (since it has the lowest probability of exhibiting an active lesion); similarly, the designation of a lesion on a pit and fissure site on a molar (or premolar) as an "active lesion" has the highest probability of being correct. Hence, by using these two sample types as the validation sites, the probability of producing both FP and FN results is minimized, albeit to a currently unknowable extent. It is, of course, possible that the buccal surface of a canine (or incisor) might be undergoing demineralization at a subclinical level but that probability is small. Equally, it is possible that a lesion designated as active by a clinician, using subjective assessment methods, may in reality have become inactive, but, by using the pit and fissure sites of molars (and premolars), that probability is minimized.
It should be noted at this point that, in terms of the degree (or level) of release of calcium ions from the enamel surface, a sound site is by definition identical to an arrested/inactive lesion site, since neither is undergoing net mineral loss, even though the macroscopic clinical appearance will be considerably different. This equivalence is key to the understanding of the methodological approach taken in this study.
To reiterate, in an ideal world, testing the performance of a new technology for its ability to determine if a lesion is active or arrested (inactive) would involve comparing lesions which are known to be definitely active with those which are known to be definitely arrested.
However, because there is as yet no identified clinical or technological (in vivo) gold standard validation method for determining lesion activity to a high degree of accuracy, this ideal comparative methodology is not possible in any meaningful sense, due to the considerable limitations of the current clinical validation techniques. The use of such a clinically designated "active" versus "inactive" lesions comparison methodology is thus totally inappropriate and would beg the question "Is the new technology more accurate than the validator?"-analysis of the results could not determine which of the techniques-test or validator-is closer to identifying the "true" activity status of each lesion.
Thus, for the current study to validate this novel bioluminescence technique based on the identification of the presence of calcium ions in the surface enamel, the authors chose to use the sampling method described above, that is, using specific tooth sites where: (a) there is a high probability that a clinically designated "active" lesion is indeed active (an identified white spot lesion in a plaque stagnation area) and (b) there is a high probability that a clinically designated sound site is indeed sound (which is equivalent, in terms of the surface enamel pore fluid concentration of calcium ions, to an arrested lesion). By this means, together with specific training of the investigators (in the identification of active lesion and sound sites) and the independent review of the study images, the authors sought to minimize the probability of validator error.
The use of a comparison of two different tooth sites raises a potential issue with examiner bias in assessing the images, since the independent (second) examiners could infer that the smooth surfaces had been classified as sound and occlusal surfaces had been assessed as exhibiting an "active lesion" by the first examiner. This potential bias issue is dealt with in Section 4.
This study was required for submission to the United States regulators and therefore the study design was agreed in advance with the US FDA prior to local approvals. It followed a smaller pilot clinical study, and the completion of toxicological and biocompatibility testing as required by the regulators.
The purpose of this current prospective, multisite, nonrandomized, post-approval, clinical study was to assess the level of agreement in carious lesion activity status between the findings of general dental practitioners' visual and tactile assessment with those of the Calcivis imaging device in two teeth populations (i.e., with and without visible lesions).The performance of the device was measured by presence or absence of elevated luminescence, determined from intraoral image mapping of teeth surfaces with or without a visible lesion. Any adverse events were also collected.

| MATERIALS AND METHODS
This was a prospective, five-investigator, nonrandomized, post-

| Participants
Eligible participants ≥6 years old had at least one unrestored, accessible, free, smooth buccal surface on a canine or incisor, away from the gingival surface, with no visible lesion (coded ICDAS 0) and/or one unrestored, accessible, erupting or erupted molar or premolar with a visible initial lesion (coded ICDAS 2 or 3) on the occlusal surface, in a plaque stagnation area. The only exclusion criteria were: pregnancy; breastfeeding; tooth bleaching within 2 weeks of imaging; ongoing remineralization treatment; fixed orthodontic appliances; and currently taking part, or within 3 months of taking part, in a clinical study.
The "within 2 weeks of bleaching" limitation was applied in order to limit any possible interference with the calcium ion dynamic equilibrium in the enamel pores resulting from any bleaching procedure, which might affect the Calcivis system.

| The CALCIVIS imaging system
The CALCIVIS imaging system consisted of an intraoral device and cradle; single-use sterile device syringes; vial adaptors; and single-use applicators (Figure 1; Perfect et al., 2012;Perfect & Longbottom, 2008 with the trainer on lesion presence/activity or sound status being required to achieve competence. Because of the generally recognized issue of relatively lower intraexaminer reproducibility regarding ICDAS code 1 lesions compared to those for code 2 and 3 lesions in this study, in order to minimize/eliminate that potential problem, the examiners were asked F I G U R E 1 The pre-commercial CALCIVIS imaging system (as used in the study) to identify ICDAS code 2 or 3 lesions when seeking active lesions to image.

| Study visits and clinical procedures
The outline of the study design is shown in the Flow chart in Figure 2.
Participants were identified by the investigators during routine dentist visits as meeting the inclusion/exclusion criteria and provided with study details. If they agreed to participate, they returned to the dental All tooth surfaces to be imaged were cleaned by the investigator by brushing with nonfluoridated water or dental paste, then rinsing with an air-water spray from a conventional dental 3-in-1 device. No specific dental paste was recommended, if used, the choice was at the discretion of each investigator. Participants then rinsed thoroughly with nonfluoridated tap water. Following thorough air-drying of each The outline of the study design, showing the flow of the patients' (Pts) experiences tooth surface to be imaged, the investigator took a color photograph with a standard intraoral camera (CareStream CS1500-factory settings) for reference and recorded the ICDAS score and lesion activity status.
In preparation for use of the CALCIVIS imaging system, care was taken to ensure the tooth surface and surrounding area were free from saliva using appropriate moisture control aids (buccal dryguards, dry tips and saliva-ejectors). Each tooth was then air-dried for 5-10 s. The investigator instigated the image-capture proce- dure. An image of the chosen tooth was first captured as a visible, black and white, light image, then the photoprotein was applied, followed by capturing the luminescence signal ( Figure 3). The system is fully automated so that both images are taken within less Visit 2 took place 7-14 days after Visit 1, when an oral examination was performed by the investigator.

| Independent investigator review
After all images were taken and verified as acceptable by each first (originating) investigator, an independent review of the images was undertaken by one of the other study investigators to assess presence or absence of elevated luminescence on the surface of the imaged teeth. The determination by each independent reviewer was made off-site with no input from the originating investigator. The reviewing investigator was provided with the laptop with the CALCIVIS imaging system images (visible, luminescence, and overlay; Figure 3) and the reference intraoral color images, along with the following information for each tooth: tooth ID and surface; image number; photographic evidence of "area of interest" (lesion or the sound area: Figure 3, column 1). The independent reviewers F I G U R E 3 Examples of black and white, luminescence and combined images of teeth with (a) an active lesion (true positive), and (b) no lesion (true negative). The pink outline in columns 1 and 3 are the area of interest (AOI) delineated by the investigator were blinded to the first (originating) investigator's recorded ICDAS scores/activity status.
The designation of which independent reviewer assessed the images from each first investigator is shown in Table 1.

| Safety
Any adverse events were recorded at both visits (v1-baseline imaging visit and V2-7 to 14 days post-imaging). If appropriate, a color image was taken of the adverse event.

| Statistical analysis
The study sample size was calculated in terms of number of teeth of each tooth population required. As no more than one tooth of each population was imaged in each participant, the number of participants required was at most the total number of teeth required. To ensure poolability of data from each investigator, up to 36 participants were recruited to the study by each of the five investigators to obtain a minimum of 17 evaluable images of each tooth population ("sound," "active lesion") per investigator, after which recruitment was stopped. is the percentage agreement in tooth status population). All statistical analysis was performed using SAS version 9.2 (SAS Institute, Inc., Marlow, UK).
CALCIVIS imaging system performance was measured by presence/absence of tooth surface luminescence, as assessed via expected level of agreement between the CALCIVIS imaging system and investigator rating of suitable teeth in two teeth populations: "sound teeth," and "teeth with active noncavitated lesions," based on previous study data and expert opinion (data on file). "Sound teeth" were expected to correspond to "no luminescence" according to the CALCIVIS imaging system in at least 70% of cases; "teeth with active lesions" were expected to correspond to "luminescence" in at least 70% of cases. CALCIVIS imaging system safety was measured by the collection of adverse events.
All statistical tests were conducted one-sided with a 2.5% level of significance; no adjustment was made for multiple testing. The percentage agreement of the first investigator's ICDAS activity score and the reviewing investigator's CALCIVIS finding ("no luminescence" or "luminescence") was presented along with an exact one-sided 97.5% confidence interval (97.5% CI) and p-value comparing the percentage agreement to 0.5 for teeth rated with "no visible lesion" that were assessed as "no luminescence" by the CALCIVIS imaging system or teeth rated with an "active lesion" that were assessed as displaying "luminescence" by the CALCIVIS imaging system.
The analysis of agreement was performed on the Agreement Population, which included all teeth on which there was an ICDAS score by the first investigator. Luminescence data was collected and analyzed from the reviewing investigator only. Missing data arising from teeth where the CALCIVIS imaging system assessment was uninterpretable for reasons unrelated to the assessment outcome were not imputed. Adverse events were summarized descriptively for the safety population, which included all those on whom the CALCIVIS imaging system was used.

| RESULTS
Of 121 eligible participants, 111 were recruited of which 110 were imaged with the CALCIVIS imaging system. There were 61 males (55.5%) and 49 females (44.5%) in the Safety population, with an age range of 7-74 years (mean 24.3 years; SD 12.22).  Table 1 shows which independent reviewer assessed the images from each original investigator.

| Safety
There were no adverse events related to either the device or photoprotein.

| DISCUSSION
The accurate assessment/identification of caries lesion activity is one of the key elements in caries lesion diagnosis and management (Pitts et al., 2017). The World Dental Federation (FDI) has, in September 2019, published a Policy Statement on Caries Lesions and First Restorative Treatment (FDI, 2019)-this seeks to "encourage a shift from a restorative approach to caries management to the delivery of preventive dental medicine." It specifically recommends that any tissue removal decision must consider both lesion stage and activity. The current conventional means of assessing lesion activity is to use sequential assessments of a detection system (visual or technology-based), such as ICDAS coding or radiography, and compare the size/depth of a lesion over time. This involves inherent disadvantages, such as difficulties in standardization issues and the relatively long timeframe needed between assessments, in addition to the risk of leaving untreated an active lesion (Pitts, 2004).
The surface of a sound enamel site is essentially "smooth" with small pores at the submicron level of imaging, whereas an active caries lesion displays a "roughened" surface with enlarged enamel "pores," termed "focal holes." An arrested lesion, as a result of a F I G U R E 4 (Continued) combination of abrasion of the roughened surface and redeposition of mineral into/onto the surface porosities of the lesion, is essentially smooth but with a slightly disrupted surface. While it is relatively straightforward for a dental professional to differentiate between a sound enamel surface and one with a distinct white spot lesion, it is difficult to differentiate objectively between the optical properties of an active and an arrested white spot lesion, with only subtle and minor differences in quantity and quality of enamel surface light reflectivity (Fejerskov & Larsen, 2015). Similarly, using a 0.5 mm ball-ended metal probe as a vibration-assessment tool in an attempt to detect micron level changes in surface morphology are inherently highly subjective. The CALCIVIS imaging system was developed to aid in the characterization step of assessment, when a potential carious lesion has already been detected, or suspected, and more information is needed about that lesion regarding its current activity status in relation to ongoing calcium ion loss. In the context of testing the CALCIVIS imaging system, a clinically categorized sound enamel surface is equivalent/identical to an arrested caries lesion. This is due to no net loss of calcium ions such that the average mineral density (or, the obverse, porosity) of the surface enamel layer at a sound site is effectively identical to that of an arrested lesion, even though the appearance can be quite different between translucent sound enamel and a white/opaque underlying (demineralized) enamel (carious) lesion (Fejerskov & Larsen, 2015). The results of the comparisons in this study showed that use of the CALCIVIS imaging system correctly identified 90.7% of lesions designated as "active" by the first investigator (true positives), with 9.3% of these lesions producing an "apparent" false negative result. The CALCIVIS imaging system correctly identified 97.8% of the sound sites (true negatives), with 2.2% of sites apparent false positive results.
Even though the CALCIVIS imaging system is not a caries detection device, few, if any, lesion detection-only techniques reach such high levels (summarized by Neuhaus & Lussi, 2018) as the above true positive and negative rates for the CALCIVIS imaging system in this study.
The apparent false negative results, where there was a lack of luminescence from sites designated as active lesions, may have been due to incorrect activity status designation by the first investigators, given the widely recognized subjective nature of the clinical-visual assessment of enamel lesion activity status. In the "real-world," the dental professional will integrate CALCIVIS imaging system test results with clinical knowledge, including patient caries risk, and interpret the results according to the total clinical picture, thus the 9% false negative figure is likely to be an overestimate of the "true" percentage false negative value and, with a risk-informed recall regime in place, will be modulated through reimaging within a risk-appropriate timeframe. In relation to the 2% false positive rate, in the context of "real-world" practice, the clinical impact of such a low false positive error will be minimal, since preventive measures for active enamel lesions are nondestructive, hence no irreversible operative treatment decision for an inactive lesion would be involved.
Theoretically, there would appear to be an issue of potential bias in the methodology, since, by comparing two different tooth surface types, the independent examiners could infer in advance the designated status F I G U R E 4 (Continued) ("sound" or "active lesion") of each surface (occlusal or free-smooth).
However, the status designation was arrived at using a CLINICAL VISUAL assessment, whereas the luminescence images are created independently of the macro-morphological characteristics of the enamel surface which are assessed when using clinical visual-tactile methods. The images in Figure 4 illustrate the striking differences between the luminescence levels obtained from the "areas of interest" in designated sound surfaces and a designated active lesion surfaces-these optical signal differences are so large that even though the examiners might be aware of the prior designated status of each surface, the visual signals are so starkly different that this would override any potential bias-see Figure 4. It is of note in Figure 4 that the "background" levels of luminescence intensity from the nonfissure aspects of the occlusal sites are generally at the same levels as those from across the whole of the freesmooth surface sites in the same subjects. The fact that several "false negative" and two "false positive" results for the status designations were actually obtained, despite this prior awareness, indicates that any potential bias effect was overridden by the compelling visual evidence of the extent of the luminescence (optical) signal.
Potential confounders of the Cis include two obvious sources of calcium ions: saliva and the dental biofilm (plaque)-these factors can be effectively controlled using standard conventional clinical techniques (of isolation, air-drying and cleaning) which form part of the Cis pre-imaging protocol.  (Young et al., 2015). CALCIVIS data, and other data, such as the patient's caries-risk category, could be combined with these scores to determine the optimum treatment option for lesions with such codes.
In conclusion, the results of this clinical study show that the CALCIVIS imaging system can be used to visualize active demineralization sites in clinically identified enamel lesions and is safe for clinical use.

CLINICAL SIGNIFICANCE
The study results indicate that the Cis provides a highly accurate one time-point aid for clinicians in determining the presence of elevated calcium ion concentrations (indicative of activity status) of clinically detected enamel caries lesions compared to healthy sites (caries inactive)-this differentiation between active and inactive sites is of critical importance in deciding the optimum treatment/therapy for individual enamel lesions.

ACKNOWLEDGEMENT
This study was funded/sponsored by CALCIVIS Ltd.

AUTHOR CONTRIBUTION
All authors contributed to the study design and interpretation and to drafting the manuscript, including final approval. NP, CL, MW and BV also contributed to the conception of the study; NS also contributed to data acquisition

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in www.clinicaltrials.gov, reference number NCT 02780856. Note: All results pooled. All lesions with ICDAS scores 2 and 3 were assessed by the originating investigators as active lesions.