A general observation from that review was that current methods tended to be more specific than sensitive (that is, their use results in relatively more false-negative findings than false-positive findings). Another review prepared for the conference examined the performance of methods currently under development.² That review concluded that several of these methods showed promise, especially in improving the sensitivity of detection, although insufficient clinical data were available to support their immediate use.

Since these reviews were prepared, several additional studies have appeared that evaluated the DIAGNOdent device, or DD (KaVo America, Lake Zurich, Ill.), a caries detection system based on laser-stimulated fluorescence that had been considered only briefly in both of the systematic reviews, because the device had just become commercially available at the time of the conference.

Clinicians are constantly seeking better diagnostic methods, and they need thorough assessments of new technologies to help guide their decisions. A summary of the literature with case descriptions recently appeared in a supplement to a dental journal,³ together with a description of how to integrate the device into hygiene practice⁴; however, an in-depth evaluation of reports of the performance of the DD has not appeared in the clinical literature. The purpose of this article is to report the results of a systematic review of the literature describing the performance of the DD.

	METHODS

TOP ABSTRACT METHODS RESULTS DISCUSSION IMPLICATIONS FOR CLINICAL... RECOMMENDATIONS FOR FURTHER... CONCLUSION REFERENCES

 
Our key clinical questions addressed the performance of the DD in detecting dental caries. Specifically, we examined performance with regard to detection of occlusal and smooth-surface caries in dentin and enamel. We specified four inclusion criteria:

– the study assessed a commercially available version of the device;

– the study used human teeth with and without carious lesions;

– the study expressed diagnostic performance in terms of sensitivity and specificity in detecting dental caries; and

– the study determined diagnostic performance by comparing it with a histologically based reference standard or gold standard. We did not set any specific criteria for the gold standard, and accepted several methods for the direct assessment of tooth tissues.

We searched the MEDLINE database in May 2003, and updated the search in June 2004 using key words and groupings ("diagnodent" or "laser fluorescence" or "fluorescence" and "dental caries"). The search was limited to English-language articles and studies involving human subjects. We searched back only to 1999 because a thorough search of prior years had identified no eligible studies.¹ The search yielded 115 citations.

Two evaluators reviewed the titles and abstracts, which resulted in the consensus exclusion of 75 studies. On review of the 40 complete studies, the evaluators excluded 15 studies because they lacked sensitivity and specificity data,^5–7 lacked a histologically based gold standard,^8–12 did not use the DD,^13–16 lacked original data (a review or case study)^17,18 or lacked noncarious teeth.¹⁹ A review of the reference lists of the 25 included publications^20–44 yielded no additional citations. We contacted the authors of one study to obtain complete sensitivity and specificity data.²⁸

A single abstractor extracted data from the 25 included studies and entered them into an evidence table (available in unedited form at "www.ada.org/prof/resources/pubs/jada/reports/caries_detection.pdf"). A second reviewer subsequently verified the information in the evidence table by comparing it to the studies themselves. Included in the abstracted information were items addressing the five essential elements of a quality assessment of diagnostic studies (that is, an appropriate study population, an adequate description of the test, an appropriate reference standard, a blinded comparison of test and reference samples and avoidance of verification bias).⁴⁵ We did not combine these quality-related items into a quality score. Rather, we discuss below the implications of failing to achieve these essential elements.

We made a single subjective assessment of the likelihood of scoring bias based on the extent to which the research design separated the diagnostic determinations for different methods and repeated applications of the same method. When a choice was available, we selected for entry into the evidence table those results in which teeth were lightly air-dried. For reliability entries, we calculated the mean intraexaminer and, where appropriate, interexaminer values from the data presented in the articles.

	RESULTS

TOP ABSTRACT METHODS RESULTS DISCUSSION IMPLICATIONS FOR CLINICAL... RECOMMENDATIONS FOR FURTHER... CONCLUSION REFERENCES

 
The 25 included studies fell into four general categories. Sixteen studies assessed the in vitro performance of the DD on the occlusal surfaces of primary and permanent posterior teeth. Four studies reported similar in vivo assessments. Two studies examined the caries-detection performance of the DD on smooth surfaces. Finally, three studies addressed special applications: the detection of secondary caries and the detection of residual dentinal caries in simulated cavities.

In vitro detection of occlusal dentinal caries. Twelve studies examined the detection of occlusal dentinal caries in permanent teeth,^{20–29,31,32} and four studies examined the detection of dentinal caries in primary teeth.^20,33–35 Table 1 (page 1416) lists the salient features of these studies and the results of the performance assessments of the DD and other caries detection methods reported for comparison purposes. The DD sensitivity scores for permanent teeth ranged from 0.19 to 1.0. Nine of the studies reported sensitivity scores in a relatively narrow range near the upper end of the performance scale (0.79 to 1.0), indicating that in these studies, the relative proportion of false-negative to true-positive determinations was small. Two of the remaining three studies reported much lower sensitivity scores (0.19 and 0.25).

The comparison data from eight studies describing the performance of visual methods for detecting occlusal dentinal caries show relatively high specificity values (from 0.68 to 1.0), but much lower and somewhat more variable sensitivity values (from 0.12 to 0.82). This pattern of performance reflects the larger group of results reported for visual detection methods in a recent systematic review of diagnostic methods.¹ In all but two of the studies included in our review, the sensitivity of the visual method was lower than that of the DD, while the specificity was higher. In three comparisons of bitewing radiographs with DD, sensitivity was lower for radiographs in all three studies,^25,34,35 while specificity was higher for radiographs in two studies.^34,35 A mixed pattern is evident in five comparisons of DD with the electrical conductance method (all studies used the Electronic Caries Monitor, or ECM, Lode Diagnostics, Groningen, Netherlands). Sensitivity was lower for ECM in four of these studies,^22,23,26,32 and specificity was higher for ECM in three studies.^22,23,32

Several differences in specific design features reduced the potential for equivalence of the studies (for example, similar populations) for purposes of cross-study comparisons of DD performance. Perhaps most importantly, the threshold for the DD score (that is, the score at which caries was identified) varied across the studies (range = >10 to >22.1 for permanent teeth; >9 to >17 for primary teeth). In some studies, the manufacturer’s suggested threshold was used; however, some studies that reported using the manufacturer’s threshold actually used different thresholds.^24,25,29,35 In other studies, the authors determined the threshold post hoc to maximize the sum of the sensitivity and specificity scores.^{22,26,31,32,34} In addition, about half of the studies used formalin as the storage medium, and one used chloramines, both of which may increase the fluorescent response.^40,46

Differences in the lesions being detected and in the detection task itself also make comparisons somewhat problematic. The prevalence of dentinal caries in the study samples varied from 16 to 52 percent, and was not reported in three studies. Although all detection took place on occlusal surfaces, most studies assessed a specific site, while three studies assessed the entire occlusal surface^27,31,35 and three studies assessed multiple sites on the same surface.^22,28,29

The histologic definition of the extent of penetration for classification as dentinal caries was essentially the same in all but one study, in which deep enamel caries was included in the definition.²⁷ However, the studies used an assortment of histologic techniques to assess the extent of penetration (see evidence table). The principal differences among methods included serial versus hemisectioning, thickness of serial sections, degree of magnification used for inspection, direct versus indirect inspection (photograph, photomicrograph) and use of stains.

Two of the studies^26,31 reported perfect DD specificity values of 1.0 (that is, no false-positive results), and two others reported values of 0.98²⁹ and 0.95.²⁸ These values are higher than those reported in the remainder of the studies (range, 0.52 to 0.92). Several aspects of these studies may help explain these higher values. In the two studies with perfect specificity (which had two investigators in common), the threshold score was calculated to maximize performance. One of these studies²⁶ included only those teeth for which the highest DD score was recorded on the exact same site at two separate assessments two weeks apart. Consequently, the study excluded from the analysis 11 percent of the originally included teeth.

In the second study with perfect specificity, the authors calculated performance using only the second reading of the DD system, even though the first reading was used to select the sites for measurement and for subsequent histologic assessment. The second reading was thought to be more representative, because there was less operator error.³¹ We noted in the discussion section of this study that "one of the 76 teeth had a rather high DD reading, but the corresponding micrograph showed no visible caries."³¹ This statement seems to contradict the reportedly perfect specificity.

The 0.98²⁹ and 0.95²⁸ specificity values for DD in the other studies perhaps are explained easily by noting the very low sensitivity values in these studies. Also, one study used a DD threshold score of 10.²⁹ Although this was the lowest number used for a DD threshold, it also was used in an earlier study³² that resulted in a sensitivity of 0.79 and a specificity of 0.84. The authors of the later study speculated that the differences may have been due to differences in calibration and in the storage medium.²⁹

In vitro detection of occlusal enamel caries. Ten studies examined the detection of enamel caries on occlusal surfaces of permanent teeth, and three studies examined the detection of enamel caries in primary teeth. Table 2 summarizes the designs and results of these studies, all but one³⁰ of which also reported results for dentinal caries detection. Six of the studies analyzed the performance of DD in detecting any carious lesions, rather than just enamel lesions.^{20,22,23,29,30} Three of these studies reported high DD sensitivity values (0.95, 0.93) accompanied by low specificity values (0.53, 0.50, 0.24) (that is, although almost all lesions were detected, large proportions of sound teeth also were indicated as being carious). The remaining three studies reported moderately high sensitivity (0.73 to 0.77), with more varied specificity (0.49 to 0.85). All but one of the remaining seven studies exhibited a pattern of low-to-moderate sensitivity (0.38 to 0.79) and moderate-to-high specificity (0.68 to 0.95), the opposite of the pattern for dentinal caries.

In vivo detection of occlusal dentinal caries. Table 3 summarizes four in vivo studies^36–39 examining the detection of occlusal caries in dentin. In three of these studies, only those teeth deemed to have dentinal caries on the basis of visual,³⁷ visual and/or radiographic,³⁸ or visual and radiographic and DD criteria³⁹ were assessed macroscopically after being opened. None of these studies reported the visual and radiographic criteria used to determine the presence of dentinal caries. The authors of the studies estimated specificity scores by assuming that all unopened teeth did not have dentinal caries. These three studies reported similarly high DD sensitivity values. Estimated specificity values were lower and more varied.

The fourth study in this group examined the accuracy of caries detection by using histologic validation immediately after the exfoliation or extraction of the primary molar.³⁶ The results of this study were quite different from those of the other studies, with lower sensitivity and higher specificity values for DD, and just the opposite pattern for bitewing radiographs. Both of these patterns are not reflected in other studies of occlusal surfaces included in this review.

In vitro detection of smooth-surface caries. Two studies assessed DD performance in detecting smooth-surface caries.^40,41 In both studies, proximal surfaces of premolars extracted for orthodontic reasons were assessed as surrogates for facial and lingual smooth surfaces. Although not acknowledged in either report, these studies seem to have used the same material, with the experimental data collected before the teeth were sectioned for histologic assessment. The teeth were stored in thymol and then in formalin for two weeks.

In the first study,⁴⁰ the same operator obtained DD scores twice (two weeks apart) while the teeth were stored in thymol, and then again after they were stored in formalin. In the second study,⁴¹ a single DD score was obtained. Thresholds for each assessment were calculated post hoc to maximize performance. Table 4 shows these thresholds and the sensitivity and specificity values. The specificity values tended to be high, while the sensitivity values were lower. The variations in the thresholds and performance values between the two thymol trials and between the two formalin trials suggest that on proximal surfaces, the reproducibility of the readings is less than that reported for occlusal surfaces.

One study⁴⁴ examined performance of the DD in detecting residual caries in dentin after removal of obviously carious dentin. The authors used DD, a visual/tactile method, visible fluorescence and dye to assess 40 teeth with proximal or smooth-surface dental caries. Twenty of these lesions had been ground to remove the covering enamel, and 20 had been ground further to eliminate the majority or all of the lesion. The visible fluorescence method, which had been used to guide the extent of grinding, exhibited the best performance (sensitivity, 0.94; specificity, 0.83). DD exhibited the second-best performance (sensitivity, 0.88; specificity, 0.70), followed by the visual/tactile method (sensitivity, 0.76; specificity, 0.65). Caries dye had the lowest sensitivity (0.65) and specificity (0.17).

Finally, three studies correlated the DD score with a measurement of the greatest histologic depth of the lesion. The mean correlation coefficients for occlusal dentinal caries were 0.38²⁶ and 0.51,²³ and for proximal caries, 0.86.⁴⁰ These results echo the variability in the sensitivity and specificity results reported above.

	DISCUSSION

TOP ABSTRACT METHODS RESULTS DISCUSSION IMPLICATIONS FOR CLINICAL... RECOMMENDATIONS FOR FURTHER... CONCLUSION REFERENCES

 
Although many systematic reviews are accompanied by meta-analyses for the purpose of calculating weighted averages of an effect or a performance estimate across studies, there are several situations in which meta-analyses are not appropriate. According to Deeks,⁴⁷ in diagnostic studies, "meta analysis should only be considered when the studies have been recruited from clinically similar populations, used comparable experimental and reference tests, and are unlikely to be biased. Even when these criteria are met, there may still be such gross heterogeneity between the results of the studies that it is inappropriate to summarize the performance of a test as a single number."

Unfortunately, none of the conditions listed above was met by the studies included in this review, and heterogeneity was present. The populations used for in vitro studies all consisted of extracted posterior teeth, but these teeth usually were of unreported provenance. The studies used differing distributions of tooth types, and selected teeth using differing—and sometimes unreported—visual selection criteria with respect to the presence and extent of caries.

The experimental test was comparable in that all studies used the DD, but the analyses used a variety of detection thresholds, some of which were calculated to maximize performance for the specific sample of teeth studied. Obviously, this practice negates generalization of the results. The reference tests used for validation varied across the studies, and ample evidence exists that reference tests do not always yield the same truth.^48–50 Owing to small sample sizes and short or unreported washout periods between assessments in many of the included studies, we cannot discount the potential for bias in many studies.

Finally, simple ² tests for homogeneity on the in vitro sensitivity and specificity scores for dentinal and enamel caries indicated there was substantial heterogeneity in all four distributions of values. For these reasons, we did not attempt to perform a meta-analysis of the performance data.

Despite the lack of a meta-estimate of performance, a pattern of results did emerge from the 12 in vitro studies assessing the performance of DD in detecting dentinal caries on occlusal surfaces of permanent teeth. The DD appears to be quite sensitive, with most studies reporting that it detected between eight and nine of every 10 histologically confirmed lesions. However, the extremely low sensitivity values reported in two studies (0.25 or less), as well as differences in sensitivities in studies using the same threshold values do raise some concerns about the degree to which sensitivity results may be affected by examiner technique.

The relatively high sensitivity of the studies is accompanied by somewhat more varied specificity, with four of the studies reporting that one-third or more of noncarious surfaces were incorrectly identified as being carious, and four other studies reporting that virtually no noncarious surfaces were incorrectly identified as being carious. Results in primary teeth exhibited a slightly different pattern, with specificity being less variable and slightly higher than sensitivity.

What is quite clear is that in the direct comparisons with visual methods reported in seven of these in vitro studies, the DD was more sensitive and less specific than was the visual method. Based on only three comparisons, two of which used primary teeth, the sensitivity for bitewing radiographs was lower than that for DD, while the specificity appears to be less variable than that for DD and in the middle of the range of specificity values reported for DD.

When we compared DD with ECM, we found a pattern similar to that for radiographs. Three of four in vivo studies^37–39 assessing the detection of occlusal caries in dentin exhibited patterns similar to those of the in vitro studies. These studies reported very high sensitivity scores (all above 0.90), but lower and more varied specificity scores. The fourth in vivo study³⁶ was an outlier in this group, with high specificity and moderate sensitivity. However, this result echoes the general differences between primary and permanent tooth patterns noted among the in vitro studies.

For the detection of occlusal enamel caries, the pattern that emerged from the in vitro studies is less definitive. In seven studies detecting caries that extended only into the enamel of permanent and primary teeth, sensitivity varied widely, and always was lower than 0.80, while specificity tended to be higher (no value below 0.68). There were too few comparisons between DD and visual and ECM methods to be informative. When both enamel and dentinal occlusal caries were detected, sensitivity tended to be higher, with widely varying specificity values.

We must note again that several aspects of these studies make any generalization of these patterns somewhat tenuous. Most of these concerns are related to the essential quality elements of diagnostic studies noted above. All of the studies did well in describing details of the test, and all but the in vivo studies avoided verification bias. However, information concerning the reference or gold standard was less complete. The extent of blinding in determinations of test and reference results varied and often was not clear, as indicated in the scoring bias column in the evidence table. Descriptions of study populations, and thus the comparability and generalizability of these populations, tended to be limited. Thus, most studies met two of five criteria,⁴⁵ and shortcomings in three criteria reflect weakness in research design or reporting.

The in vivo studies that depended on visual validation on opening the fissure were the most problematic in meeting the essential quality criteria. The sensitivity scores for one or more of the diagnostic methods used may have been biased upward, because the calculation of sensitivity was based only on visual verification from that subset of teeth not likely to be carious, and also not likely to have a subtle lesion (since both visual and radiographic methods tend to have relatively high specificity and relatively low sensitivity).

All of the specificity calculations in these studies were estimates, based on the assumption that unopened teeth are not carious. Thus, although false-positive findings may have been detected in the subset of teeth that were operated on, specificity (that is, the rate of false-positive findings) must be estimated, rather than calculated, because the number of histologically accurate true-negative findings cannot be determined. Because estimated specificity scores assume perfect specificity with regard to the criteria used to determine which teeth are opened, they may overestimate true specificity.

Finally, in all three studies, the decision to open a tooth was made by the same examiner who collected the diagnostic data and performed the histologically based assessment of the extent of caries. Such unblinded designs are subject to bias, both in the selection of teeth to be operated on and in the determination of the presence of caries after the fissure is opened.

The two studies of diagnostic performance on smooth surfaces offer only limited information. These two studies are not independent because they used the same sample. They are not generalizable because they assessed caries on proximal surfaces typically inaccessible to the DD. No information is available to predict how the device will perform on facial and lingual smooth surfaces. Interestingly, the variations in performance between the two thymol applications do not suggest the same degree of test-retest reliability reported in the occlusal studies. The lack of performance differences between the two formalin applications is puzzling, because the threshold score was different for these two applications.

The studies of DD performance in detecting marginal caries associated with restorations, as well as residual dentinal caries after tooth preparation, must be regarded as initial explorations of these topics. Performance was good in terms of detecting marginal caries, and better than that of visual detection, but the illogical reversal of the threshold score in one study (that is, a higher threshold score for enamel caries than for dentinal caries) does not inspire confidence in these results. The simulated nature of residual caries limits the conclusions that can be drawn from this assessment.

	IMPLICATIONS FOR CLINICAL PRACTICE

TOP ABSTRACT METHODS RESULTS DISCUSSION IMPLICATIONS FOR CLINICAL... RECOMMENDATIONS FOR FURTHER... CONCLUSION REFERENCES

 
From a clinical perspective, any diagnostic method that improves on the performance of current methods for detecting occlusal caries merits consideration. However, it is important to consider the aspect of diagnostic performance that is improved. A recent systematic review demonstrated that visual and visual/tactile methods may offer the highest and least variable diagnostic specificity, while electrical conductance methods may elicit the highest and least variable diagnostic sensitivity.¹ Even though dentistry always has prized high diagnostic sensitivity (that is, the ability to find disease), high, predictable diagnostic specificity (the ability to rule out disease) has grown in importance, because epidemiologic caries prevalence rates have fallen and caries progression rates have slowed.

For a given diagnostic method, as more and more teeth are found not to be carious (true-negative findings), the number of teeth incorrectly deemed to be carious (false-positive findings) will increase. For example, if a diagnostic method has a sensitivity of 0.90 and a specificity of 0.90, an equal number of false-positive and false-negative findings will occur during examination of occlusal surfaces of teeth for which the caries prevalence is 50 percent. However, among a population of surfaces in which the caries prevalence is 10 percent, nine false-positive diagnoses will be made for every false-negative diagnosis.

In the clinical setting, a false-positive diagnosis of dentinal caries usually will lead to some form of tooth preparation with irreversible iatrogenic damage, while a false-negative diagnosis may have no long-lasting effects, depending on the frequency of examination and the rate of caries progression. The results of a recent study of early treatment of questionable carious lesions versus watchful waiting of suspicious areas suggest that delaying operative intervention for up to two years did not result in the need for excavation of larger amounts of dentin than did immediate intervention.⁵¹

We find it discouraging that although the pattern of results identified in this review of the DD suggests that it offers some measure of improved diagnostic sensitivity in detecting enamel and dentinal caries on occlusal surfaces, the improved sensitivity comes at the cost of specificity values that are lower than those achieved with visual examination. For this reason alone, clinicians probably should not rely on the DD as their primary diagnostic method. In addition, uncertainty exists over what diagnostic threshold value to use, the in vivo evaluation of the method has been less than comprehensive, and the nagging problem remains of not knowing the activity status of the carious lesions that are detected.

Only one study attempted to differentiate the diagnostic performance on active and inactive lesions.³⁷ Among visually identified active and inactive enamel lesions, the difference in mean DD scores was small and score distributions were similar. It is probable, albeit not yet demonstrated, that some of the lesions identified via DD but not via visual methods and subsequently validated through histologic examination were inactive or arrested caries. We can debate whether operative intervention is warranted for arrested lesions in dentin, especially if the extent of demineralization is small. The authors of the studies reviewed here supported, to some extent, these observations. As the evidence table indicates, most investigators, in summing up the clinical usefulness of the DD, either suggested that it be used as an adjunctive diagnostic method or expressed concern about its specificity values.

Nevertheless, it appears that the DD could have several uses, such as refining a questionable diagnosis, directing preventive interventions or monitoring a suspected lesion over time. First, clinicians could use DD subsequent to the traditional visual/tactile examination to refine a questionable diagnosis of caries on the occlusal surface of a tooth. Presumably, it would be used as a "serial diagnostic test" in instances in which the results of the visual examination are equivocal.⁵² In such instances, one would assume that the dentist’s assessment of the probability of caries is 50 percent, because he or she is unable to definitively rule in or rule out caries based on the visual examination. Thus, the prior probability of caries before DD application is 50 percent.

Application of the DD then would provide additional information about the chances (the posterior probability) of disease on the surface in question. For purposes of illustration, assume that the sensitivity and specificity of the DD are 0.74 and 0.81, respectively, for detection of dentinal caries (the raw means of the 12 studies from Table 1 for permanent teeth). If the DD yields a positive finding for dentinal caries in a suspicious area (that is, the prior probability is 50 percent), the probability of a suspicious area actually having the indicated demineralization in dentin would be 80 percent. A positive result could trigger either a decision to operate or to apply fluoride, reassess the surface and compare DD scores at some time in the future.

However, this illustration assumes that the proportion of sites with dentinal caries among the suspicious areas identified (that is, the prior probability of caries) is actually 50 percent. If, for example, a dentist routinely tested all surfaces that had any stain in fissured areas, so that the prior probability of dentinal caries was perhaps only 20 percent,²⁰ the posterior probability associated with a DD score indicating caries into dentin would be 49 percent, a probability that most likely would not merit intervention.

Second, the DD may help direct preventive treatment because of its reasonably high sensitivity in identifying surfaces with demineralization. A DD test result that indicates the presence of caries would be a signal to the practitioner to consider preventive interventions, such as application of fluoride varnish or sealants. While false-positive diagnoses would be more numerous than false-negative diagnoses, the consequences of unnecessary preventive intervention are much less severe than those of unnecessary operative intervention.

Third, the DD might be used to monitor lesion activity over time. This application depends on the assumption that the DD score will increase as the lesion deepens. Under this assumption, an increase in the DD score over time would trigger operative intervention. This assumption is reflected in the concept of different diagnostic thresholds for enamel and dentinal caries, and it is intuitively appealing. Unfortunately, it is not well-supported by the studies in this review.

	RECOMMENDATIONS FOR FURTHER RESEARCH

TOP ABSTRACT METHODS RESULTS DISCUSSION IMPLICATIONS FOR CLINICAL... RECOMMENDATIONS FOR FURTHER... CONCLUSION REFERENCES

 
The deficiencies in these clinical studies form the basis for many recommendations for further research. Study procedures need to be more completely and clearly reported. In the studies in our review, essential information either was not reported or was described in such a manner that the abstractors could not agree on its interpretation. Investigators should pay greater attention to minimizing scoring bias. Establishing long washout periods between assessments of the same teeth and using multiple examiners (or perhaps even different examiners for each method) would strengthen protection against bias.

Studies should attempt to replicate the clinical task (that is, the entire occlusal surface should be evaluated). This may necessitate a rapid assessment method for histologic screening of multiple sections. Additional in vitro studies—especially those that calculate diagnostic thresholds post hoc to maximize performance measures—probably are not a good use of limited research resources. However, if additional in vitro studies are conducted, they should use the thresholds established by the manufacturer and attempt to understand the reasons for the variability in specificity rates.

Future studies should focus on in vivo assessments with histologic validation. Two types of studies are needed: cross-sectional assessments of diagnostic performance and longitudinal examinations of the association between changes in DD scores and caries penetration.

The challenge for researchers is making a sufficient number of observations on teeth that will be extracted. Third molars and premolars removed for orthodontic reasons are obvious targets, but they do not constitute a representative sample. In a large dental practice or clinic, it may be possible to identify other scheduled procedures that allow for a DD assessment and subsequent histologic assessment. Extractions performed for periodontal reasons should offer opportunities for assessment of smooth-surface lesions, and crown preparations may offer opportunities for admittedly less-precise assessment of the extent of demineralization on both occlusal and smooth surfaces.

	CONCLUSION

TOP ABSTRACT METHODS RESULTS DISCUSSION IMPLICATIONS FOR CLINICAL... RECOMMENDATIONS FOR FURTHER... CONCLUSION REFERENCES

 
This review indicates that in the available studies, the DD tended to be more sensitive than the visual method of detecting occlusal dentinal caries (that is, it identified a larger proportion of true lesions). However, the method was less specific in that the DD also identified a larger proportion of sound sites as being carious (that is, false-positive diagnoses) than did the visual method. DD performance tended to be less sensitive and more specific in detecting enamel lesions than it was for dentinal caries detection. The lack of specificity, together with the absence of a single diagnostic threshold, suggest that the DD should not be relied on as a clinician’s primary diagnostic method. Further assessment of the DD in clinical, in vivo applications to detect caries activity or progression is needed.