586763

research-article2015
JDRXXX10.1177/0022034515586763Journal of Dental ResearchVisual Inspection for Caries Detection

Clinical Review
Journal of Dental Research
1­–10
Visual Inspection for Caries Detection: © International & American Associations
for Dental Research 2015

A Systematic Review and Meta-analysis Reprints and permissions:

sagepub.com/journalsPermissions.nav
DOI: 10.1177/0022034515586763
jdr.sagepub.com

T. Gimenez1, C. Piovesan1,2, M.M. Braga1, D.P. Raggio1, C. Deery3,

D.N. Ricketts4, K.R. Ekstrand5, and F.M. Mendes1

Abstract
We aimed to perform a systematic review including a meta-analysis to evaluate the overall accuracy of visual methods for detecting
carious lesions and to identify possible sources of heterogeneity among the studies included. Two reviewers searched PubMed, Embase,
Scopus, and other sources through July 2014 to identify published and nonpublished studies in English. Studies of visual inspection were
included that 1) assessed accuracy of the method in detecting caries lesions; 2) were performed on occlusal, proximal, or free smooth
surfaces in primary or permanent teeth; 3) had a reference standard; and 4) reported sufficient data about sample size and accuracy
of methods. The data were used to calculate the pooled sensitivity, specificity, diagnostic odds ratio, and summary receiver operating
characteristics curve. Heterogeneity of the studies was also assessed. A total of 102 manuscripts (from 5,808 articles initially identified)
and 1 abstract (from 168) met the inclusion criteria. In general, the analysis demonstrated that the visual method had good accuracy
for detecting caries lesions. Although laboratory and clinical studies have presented similar accuracy, clinically obtained specificity was
higher. We also observed moderate to high heterogeneity and evidence of publication bias in most papers. Moreover, studies employing
widely recognized visual scoring systems presented significantly better accuracy as compared to studies that used their own criteria.
In conclusion, visual caries detection method has good overall performance. Furthermore, although the identified studies had high
heterogeneity and risk of bias, the use of detailed and validated indices seems to improve the accuracy of the method.

Keywords: dental caries, performance, oral diagnosis, sensitivity and specificity, ROC curve, evidence-based dentistry

Introduction visual inspection in the detection of caries lesions and explored

Caries detection in daily clinical practice is primarily performed
by visual inspection (Bader et al. 2002) because it is a simple
technique with no additional costs. Nevertheless, visual exami-
nation has some shortcomings, mainly related to its subjective
nature (Braga et al. 2010), since examiners can demonstrate
inconsistency in the interpretation of clinical characteristics of
carious lesions (Bader et al. 2002).
Several studies have evaluated the performance of visual
inspection in detecting carious lesions, and the range of
reported results has been extensive and contradictory. This dis-
crepancy may in part be due to the wide assortment of different
classification criteria used for visual inspection and the varia-
tion in the conditions in which such examinations are carried
out (Ismail 2004). To overcome these limitations, there has
been a move to develop validated caries detection systems
(Ekstrand et al. 1997; Nyvad et al. 1999; Fyffe et al. 2000;
Ismail et al. 2007). As a result, studies are necessary to investi-
gate how accurate visual inspection really is.
Systematic reviews provide the best evidence on the effec-
tiveness of a procedure and permit investigation of factors that
may influence the performance of a method. To the best of our
knowledge, no previous studies have performed a systematic
review and meta-analysis to evaluate the overall accuracy of
its possible sources of heterogeneity.
Therefore, we aimed to carry out a systematic review includ-
ing a meta-analysis to determine the overall diagnostic accuracy
of visual detection for dental caries in primary and permanent
teeth. Further, we investigated if the utilization of validated scor-
ing systems could improve the performance of the visual method.
The presence of publication bias and other possible sources of
heterogeneity—such as study setting, type of reference standard
1
Department of Pediatric Dentistry, School of Dentistry, University of
São Paulo, São Paulo, Brazil
2
School of Dentistry, Centro Universitário Franciscano, Santa Maria,
Brazil
3
School of Clinical Dentistry, University of Sheffield, Sheffield, UK
4
Dundee Dental Hospital and School, University of Dundee, Dundee, UK
5
Section of Cariology & Endodontics, Section of Pediatric Dentistry
& Clinical Genetics, School of Dentistry, Faculty of Health Sciences,
University of Copenhagen, Copenhagen, Denmark
A supplemental appendix to this article is published electronically only at
http://jdr.sagepub.com/supplemental.
Corresponding Author:
F.M. Mendes, Faculdade de Odontologia da Universidade de São Paulo,
Av Lineu Prestes, 2227, São Paulo, 05508-000, SP, Brazil.
Email: fmmendes@usp.br

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015

2 Journal of Dental Research 
method used, examiners’ experience, and type of teeth exam-
ined—were investigated as secondary objectives.
Materials and Methods
The PRISMA guideline (i.e., Preferred Reporting Items for
Systematic Reviews and Meta-analyses; Moher et al. 2009)
was followed to report this review, which is registered at
PROSPERO platform (CRD42013003718).
Information Sources
Articles that reported accuracy of detecting caries lesions by
visual inspection published until July 31, 2014, were searched
in PubMed, Embase, and Scopus databases. Unpublished lit-
erature was traced through OpenSIGLE, annals of IADR/
AADR (International and American Associations for Dental
Research), and ORCA (European Organisation for Caries
Research) congresses from 2003 until 2014. References of
included articles were checked manually.
Search
The search of electronic databases was based on an optimal
search strategy for diagnostic studies (Deville et al. 2000),
associated with the clinical situation (carious lesions) under
investigation and the detection method (visual inspection). The
entire search strategy is shown in a previous publication
(Gimenez et al. 2015).
Study Selection and Eligibility Criteria
All titles and abstracts of studies found were firstly assessed
against the inclusion criteria: 1) have some mention of visual
inspection for detection of primary coronal caries lesions; 2)
have been performed with primary or permanent human teeth,
either laboratory or clinical, and on smooth, proximal, or occlu-
sal surfaces; and 3) have been written in English language.
Full papers of included studies were then read to ensure that
they presented a clearly defined reference standard and that
they reported absolute numbers of true positives, false posi-
tives, true negatives, and false negatives or presented sufficient
data to calculate these figures. Histologic evaluation, operative
intervention, direct visual inspection after temporary tooth
separation, and radiography were the reference methods con-
sidered appropriate.
Study selection was performed independently by 2 review-
ers (T.G. and C.P.). Disagreements were resolved by discussion
with a third researcher (F.M.M.). For studies with the same
data set, only the study considered most complete was included
in this review. Articles were excluded that reported detection
performance using artificial caries, root caries, or caries adja-
cent to restorations.
Data Collection Process
Two reviewers (T.G. and M.M.B.) collected the data from
selected papers onto structured tables. Kappa values between
Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.
© International & American Associations for Dental Research 2015
examiners ranged from 0.83 to 1.00, depending on the categor-
ical variable collected. Disagreement for numerical variables
collected occurred in 6% of cases. Discrepancies were resolved
by consensus.
The following information was extracted: reference stan-
dard method, setting (clinical or laboratory), type of teeth (pri-
mary or permanent), surface evaluated (smooth, proximal, or
occlusal), sample size, examiners’ experience (nonreported,
experienced, intermediate with or without training, and nov-
ices with or without training), and accuracy data (sensitivity
and specificity). Undergraduate students were considered nov-
ice examiners, and dental practitioners were considered to
have an intermediate level of experience. We also recorded
which visual scoring system the authors used in their research.
Based on the literature, an inventory of possible scoring/clas-
sification systems was created: no scoring system reported,
authors’ own criteria, International Caries Detection and
Assessment System (ICDAS; Ismail et al. 2007), Dundee
Selectable Threshold Method (Fyffe et al. 2000), World Health
Organization criteria (1997), British Association for the Study
of Community Dentistry system (Pitts et al. 1997), Universal
Visual Scoring System (Kuhnisch et al. 2009), ERK (Ekstrand
et al. 1997), and those systems described by Nyvad et al.
(1999), Nytun et al. (1992), Lussi (1993), Downer (1975), and
Marthaler (1966). Studies that did not report any criteria were
classified as “with no criteria.” If authors used criteria with no
reference to previously published studies, we classified the
study as using their “own criteria.”
Values of true positives, true negatives, false positives, and
false negatives were also recorded when available, or the numbers
were derived from sample size, caries prevalence observed at the
sample, and sensitivity and specificity values reported by studies.
If the study had evaluated performance of the method with >1
examiner, we considered the values   of the first examiner. If the
same study had evaluated accuracy of visual examination in dif-
ferent subgroups, data were recorded as different studies.
Risk of Bias of Individual Studies
The QUADAS-2 checklist (i.e., quality assessment of studies
of diagnostic performance included in systematic reviews) was
used to assess the risk of bias of the included studies (Whiting
et al. 2011). Data were assessed and collected by 2 reviewers
(T.G. and M.M.B.).
Summary Measures and Synthesis of Results
Statistical analyses were performed separately at 2 thresholds
according to the reference standard assessment: initial caries
lesions (all lesions, independent of lesion depth or dental sur-
face integrity) and more advanced caries lesions (including
only lesions into dentin when lesion depth was assessed or
cavitated lesions when surface integrity was evaluated).
Analyses were performed in subgroups combining different
types of teeth, dental surfaces, and study setting.
First, we performed a qualitative description of included
studies and reported their results of sensitivity and specificity
Visual Inspection for Caries Detection 3
in a “paired forest plot” (RevMan 5.2, Cochrane Collaboration,
Copenhagen, Denmark; Macaskill et al. 2010).
Statistical pooling of sensitivity, specificity, diagnostic odds
ratio (DOR), and positive and negative likelihood ratios was
carried out using the DerSimonian Laird method (random
effects meta-analysis model), considering different types of set-
tings, teeth, and dental surfaces examined. Therefore, 8 analy-
ses were performed on the basis of the subgroups: clinical and
laboratory studies in occlusal surfaces of permanent teeth, prox-
imal surfaces of permanent teeth, occlusal surfaces of primary
teeth, and proximal surfaces of primary teeth. Additionally, we
summarized these numbers in receiver operating characteristic
curves based on the same subgroups (MetaDisc 1.4 Software,
Unidad de Bioestadistica Clínica del Hospital Ramón y Cajal,
Madrid, Spain). Within these groups, the areas under the curve
of summary receiver operating characteristic analysis provided
a more adequate description of study results.
Publication bias was checked through funnel plots based on
the DORs of the studies and their 95% confidence intervals (95%
CIs). Points located at the top of the graph represent studies with
a larger sample, while smaller studies are located in the lower
region. Points should be evenly distributed on both sides of the
chart, thus resembling an inverted funnel. In the presence of pub-
lication bias, the inverted funnel is asymmetrical. Two examiners
performed this assessment and reached a consensus regarding
presence or absence of evidence of publication bias. Egger’s sta-
tistical test was also performed (Comprehensive Meta-analysis
Software, Statistical Solutions, Boston, MA, USA).
The presence of heterogeneity was analyzed via inconsis-
tency (I2) based on DORs of included studies (MetaDisc 1.4).
Possible sources of heterogeneity were investigated using meta
regression analysis.
We carried out metaregression analyses to compare the
effect of methodological differences related to different vari-
ables: primary or permanent teeth, clinical or laboratory stud-
ies, type of reference standard method used (histologic,
operative intervention, or others [tooth separation, radiogra-
phy, etc.]), examiners’ experience (experienced, intermediate,
novice, and nonreported), and if authors had used validated
visual scoring systems or not. We identified the following sys-
tems that were compared with the reference category: ERK
criteria, ICDAS, World Health Organization, Nyvad scoring
system, Nytun criteria, or other visual scoring systems (i.e.,
Dundee Selectable Threshold Method, Universal Visual
Scoring System, British Association for the Study of Community
Dentistry system, and criteria described by Downer, Marthaler,
or Lussi). The last criterion systems were grouped because of
the small number of studies that used them. Studies using no
criteria or their own criteria were grouped and considered as a
reference category for the analysis. Then, we made compari-
sons with studies that used validated criteria. For metaregres-
sion analysis, DOR values were used as the outcome measure.
Relative diagnostic odds ratio (RDOR) values and 95% CIs in
relation to the reference category of the independent variables
were calculated for each condition to estimate the effect of
each variable on the accuracy of the method; statistical signifi-
cance was set at P < 0.05.
Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.
© International & American Associations for Dental Research 2015
We also performed a sensitivity analysis with the exclusion
of each study sequentially. If there was a significant change in
pooled DOR, data were presented with and without this study.
Results
Study Selection
Study selection flow diagram is shown in Figure 1. PubMed,
Embase, and Scopus searches yielded 7,851 studies. After the
removal of duplicates, the databases identified 5,808 unique
studies. Then, 5,344 studies were excluded after review of the
titles and abstracts. Following exclusions from remaining 464
manuscripts (reasons detailed in Fig. 1), 102 papers were
included. A search of OpenSIGLE and abstracts from the
annals of IADR and ORCA congresses yielded 478 investiga-
tions (Fig. 1), but only 1 was included. Others were excluded
mainly due to lack of full data about accuracy.
Study Characteristics
Publication year of included studies ranged from 1975 to 2014.
Thirty-six percent of studies were conducted by experienced
examiners; 77% were laboratory based; 78% were on occlusal
surfaces; 68% were in permanent teeth; and 80% used a histo-
logic reference standard. Characteristics of each study are pro-
vided in Appendix Table 1, with a summary of characteristics
in the footnote.
Sensitivity and Specificity of Individual Studies
Paired forest plots summarizing sensitivities and specificities
of the studies and their 95% CIs can be seen in Appendix
Figure 1a and 1b (for initial caries lesions) and Appendix
Figure 2a and 2b (more advanced caries lesions). Studies were
grouped into permanent or primary teeth, dental surface tested,
and reference standard method used. There was an observed
tendency to higher specificity than sensitivity values in the
detection of initial and more advanced caries lesions.
Synthesis of Overall Accuracy Results
Pooled sensitivity, specificity, DOR, positive and negative
likelihood ratios, I2, and summary receiver operating charac-
teristics curves were calculated separately for combinations of
tooth type, dental surface, and study setting. Data concerning
smooth surfaces were not pooled due to the limited number of
studies. Results are presented for permanent (Fig. 2) and pri-
mary (Fig. 3) teeth.
Analysis considering areas under the curve and DOR values
showed that visual inspection had similar accuracy for all types
of tooth and surfaces. A tendency toward higher pooled speci-
ficity than pooled sensitivity could be observed—except for
the initial lesions threshold on occlusal surfaces of permanent
teeth in the laboratory setting and on occlusal surfaces of pri-
mary teeth in the clinical and laboratory settings, where a
higher sensitivity value was found (Figs. 2, 3).
4 Journal of Dental Research 

Figure 1.  Flow diagram with the information through the phases of studies selection.

In general, although overall accuracy measured by areas
under the curve and DOR values was similar among studies
made in the clinical and laboratory settings, we observed a
trend of lower sensitivities and higher specificities when the
studies were performed in the clinical setting compared to the
laboratory (Figs. 2, 3). Considering the 95% CI values, we
observed that laboratory studies presented higher sensitivity
values than those of clinical studies in the following condi-
tions: more advanced occlusal caries lesions, initial and more
advanced proximal caries lesions, all in permanent teeth (Fig.
2), together with more advanced occlusal caries lesions of pri-
mary teeth (Fig. 3). For specificity, clinical studies presented
significantly higher figures for all conditions in permanent
teeth (Fig. 2) and for more advanced proximal caries lesions in
primary teeth (Fig. 3).
Studies generally presented heterogeneity varying from mod-
erate to high. Regarding the occlusal surface of permanent teeth,
values of I2 were high at the initial caries threshold (>64%) and
moderate to high (51% to 57%) at the more advanced lesion
threshold (Fig. 2). With regard to occlusal surfaces of primary
teeth, I2 values varied from 53% to 69% (moderate to high het-
erogeneity), except for laboratory studies performed in primary
teeth to detect more advanced caries lesions, which showed low
inconsistency (Fig. 3). Regarding proximal surfaces of perma-
nent and primary teeth, the method showed high inconsistency
(63% to 96%), except for clinical studies carried out to detect
more advanced caries lesions in both primary and permanent
teeth, which showed low heterogeneity (Figs. 2, 3).
Risk of Bias within Studies
The overview of QUADAS-2 graphs for laboratory and clini-
cal studies demonstrated differences in the risk of bias (Fig. 4).
Many laboratory studies present with a high risk of sample

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015

Visual Inspection for Caries Detection 5

Figure 2.  Summary receiver operating characteristic curves and synthesis of the results obtained with studies of accuracy performed in occlusal and
proximal surfaces of permanent teeth. Each study is represented by a red dot, and the pooled result of all studies by means of the receiver operating
characteristic curve is represented by the central curve, while the confidence interval is represented by the upper and lower lines. AUC, area under
curve.

selection bias. The same pattern was seen for clinical studies
but was less pronounced. Moreover, while nearly 50% of labo-
ratory studies presented high or unclear risk of bias regarding
the reference standard, these figures increased to approxi-
mately 70% in clinical studies. Yet, most laboratory studies
revealed a low risk of bias in application of the method, timing
of exams, and flow of patients or teeth, while 50% of clinical
studies present a high risk on this last topic.
Most studies (laboratory and clinical) raised concerns
regarding the applicability, when considering sample selection,
as they did not clearly indicate if the spectrum of caries present
in the study sample matched the expected prevalence in the
target population. Concerns were raised in almost 50% of labo-
ratory and 70% of clinical studies, since the target condition as
defined by the reference standard did not match the review
question. However, conduct and interpretation of the index test
in all studies did not differ from the review question indicating
that there were no concerns considering this topic. The indi-
vidual classification of each included paper regarding the
QUADAS-2 is presented in Appendix Table 2.

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015

6 Journal of Dental Research 

Figure 3.  Summary receiver operating characteristic curves and synthesis of the results obtained with studies of accuracy performed in occlusal and
proximal surfaces of primary teeth. Each study is represented by a red dot, and the pooled result of all studies by means of the receiver operating
characteristic curve is represented by the central curve, while the confidence interval is represented by the upper and lower lines. AUC, area under
curve.

Evidence of Publication Bias among the Studies
Funnel plots were performed for each type of tooth and tooth
surface at each lesion severity threshold and, separately, for lab-
oratory and clinical studies (Appendix Figs. 3–6). These demon-
strate evidence of publication bias in the following: occlusal
surfaces of permanent (Appendix Fig. 3) and primary teeth at the
more advanced carious lesion threshold in clinical and labora-
tory studies and for occlusal surfaces of primary teeth at the ini-
tial carious lesion threshold for clinical and laboratory studies
(Appendix Fig. 5). However, Egger’s test showed statistical sig-
nificance on occlusal surfaces of primary teeth in the clinical
setting at both thresholds: on the occlusal surfaces of permanent
teeth in the laboratory setting at initial carious lesions threshold
and in clinical and laboratory settings at more advanced carious
lesions threshold. Considering proximal surfaces, no statistical
significance were found. However, lack of significance should
be interpreted with caution, as a minimum of 10 studies would
be necessary in the meta-analysis to have adequate power to
detect a real asymmetry (Sterne et al. 2011).

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015

Visual Inspection for Caries Detection 7

Figure 4.  Graphic summary of risk of bias of clinical and laboratory studies based on the QUADAS-2 checklist (i.e., quality assessment of studies of
diagnostic performance included in systematic reviews).

Sensitivity and Metaregression Analysis
The sensitivity analysis did not detect any relevant change
when any study was excluded.
Regarding the metaregression results about reference stan-
dard at the initial carious lesion threshold, studies about visual
examination of occlusal surfaces in permanent teeth (38 stud-
ies) that used operative intervention as the reference standard
method demonstrated a statistically better accuracy (RDOR =
4.83; 95% CI = 1.35 to 17.34) than that of studies using histo-
logic examination (reference category), but other types of ref-
erence standard did not present any statistically significant
differences (RDOR = 3.20; 95% CI = 0.45 to 22.90). At the
initial caries lesions threshold, for occlusal surfaces of perma-
nent teeth (38 studies), examiners with intermediate experi-
ence presented poorer accuracy (RDOR = 0.21; 95% CI = 0.07
to 0.61) compared to studies using experienced examiners (ref-
erence) as did studies that did not report the examiners’ experi-
ence (RDOR = 0.28; 95% CI = 0.09 to 0.84); those using
novice trained examiners did not present a significant differ-
ence (RDOR = 1.30; 95% CI = 0.05 to 30.98).
When studies that did not report any scoring system or that
used their own criteria were compared with studies using the
ICDAS, the latter presented better performance in detecting
initial and advanced occlusal carious lesions of primary teeth
and advanced carious lesions in occlusal surfaces of permanent
teeth (Table). Studies using the ERK and Nyvad systems also
presented superiority in some conditions (Table). Other metare-
gression analyses did not present any statistically significant
differences; therefore, the data were not presented.
Discussion
Health care professionals face significant demands when trying
to keep up-to-date with the literature; issues such as lack of
access to and time taken to keep abreast of the increasing scien-
tific literature are some of the factors that prevent professionals

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015

8 Journal of Dental Research 

Table.  Metaregression Analysis to Compare the Effect of Differences of the Studies Regarding the Utilization of Visual Scoring Systems.

Permanent Teeth Primary Teeth

Occlusal Advanced Caries Occlusal Initial Caries Occlusal Advanced Caries

Visual Scoring System Lesions (62 Studies) Lesions (28 Studies) Lesions (32 Studies)

Nonreported or own criteria 1.00 1.00 1.00

ERK 2.17a (1.05 to 4.48) 2.93a (1.41 to 6.09) 1.90 (0.56 to 6.45)
ICDAS 3.45a (1.54 to 7.72) 4.86a (2.76 to 8.54) 4.72a (1.51 to 14.71)
WHO 2.45 (0.28 to 21.01) 0.80 (0.38 to 1.70) —
Nyvad — 8.82a (3.10 to 25.05) 3.60 (0.79 to 16.48)
Nytun 0.26 (0.02 to 3.10) — 1.38 (0.44 to 4.39)
Othersb 2.09 (0.66 to 6.58) — 12.41 (0.48 to 322.79)

Values are presented as relative diagnostic odds ratio (95% confidence interval). Dashes (—) indicate that visual scoring system is not included in the
analysis due to lack of studies for that condition.
ERK, Ekstrand’s system; ICDAS, International Caries Detection and Assessment System; WHO, World Health Organization.
a
Statistically significant differences compared to the reference (P < 0.05).
b
Other visual scoring systems: Universal Visual Scoring System, British Association for the Study of Community Dentistry system, and criteria
described by Downer (1975), Marthaler (1966), or Lussi (1993).

from implementing the best evidence into clinical practice
(Grimes and Schulz 2002). Hence, systematic reviews are essen-
tial to distill the available literature. A previous systematic review
examining visual inspection for caries detection did not perform
a meta-analysis (Bader et al. 2002). Since that review, potentially
significant advances in visual detection of carious lesions have
been made (Pitts 2004; Ismail et al. 2007; Kuhnisch et al. 2009);
as such, there was a need for an updated systematic review,
including, for the first time, meta-analysis and metaregressions.
The most commonly used indicators of diagnostic perfor-
mance are sensitivity and specificity. A high sensitivity is nor-
mally obtained at the expense of reduced specificity, a situation
that could lead to an increase in the number of false-positive
caries diagnoses, which in turn could lead to overtreatment of a
generally slow-progressing disease. Thus, it is more appropriate
for a method of caries detection to have a high specificity even
at the expense of a small reduction in sensitivity (Downer 1989).
This was observed for the visual inspection in this systematic
review, since we found a trend of pooled specificity being greater
than the pooled sensitivity in the majority of the analyses.
Another important point concerns differences among studies
performed under laboratory and clinical conditions. Although
overall accuracy was similar, we observed that clinical studies
presented higher specificities and lower sensitivities than those
of laboratory studies. For proximal surfaces, the difficulties in
simulating the proximal contact points under laboratory condi-
tions could be a possible explanation (Braga et al. 2009).
Moreover, the presence of bacterial plaque, acquired pellicle,
saliva, and soft tissues in clinical studies could reduce the detec-
tion of caries lesions, leading to lower sensitivity values.
Considering DOR, which is a parameter that combines detec-
tion values   of accuracy in a single parameter and does not suffer
from the influence of the threshold effect among studies, we
observed similar accuracy in all subgroups. A trend for better
accuracy in detecting initial occlusal lesions than advanced
lesions can be seen, and this may be explained by the reference
standard used in the studies, since visual examination slightly
underestimates deep lesions compared to histologic examination.
In this study, visual inspection presented an overall accuracy
similar to that previously reported with radiography (Bader et al.
2002) and fluorescence-based methods (Gimenez et al. 2013).
These adjunct methods tended to present higher sensitivities and
lower specificities; consequently, the use of visual inspection
alone seems to be effective enough for caries detection, confirm-
ing the findings of primary clinical studies (Baelum et al. 2012;
Mendes et al. 2012).
Regarding the risk of bias, the main concern is that most
studies present a high risk of bias in sample selection and raise
concerns regarding the applicability of research due to a sam-
ple not being representative. Authors should be careful in
selecting patients and teeth that represent an adequate spec-
trum of disease in the target population. Most studies of caries
detection (Gimenez et al. 2013; Gimenez et al. 2015) did not
employ a sample with an adequate spectrum; hence, this affects
the transferability of results to clinical practice. More detailed
analysis of methodological quality and clinical relevance is
found elsewhere (Gimenez et al. 2015).
Exclusion of non-English-language articles could be seen
as a possible limitation of this review. However, the exclusion
of articles published in other languages does not seem to bias
systematic reviews (Moher et al. 2000; Juni et al. 2002).
Another possible limitation was to consider only the results of
the first examiner in each study. This strategy was based on a
previous medical systematic review (Nelemans et al. 2000),
and it aimed to prevent duplication of sample data and not to
privilege any examiner, therefore avoiding under- or overesti-
mate of the accuracy.
Even when the results from laboratory and clinical studies
were analyzed separately, the heterogeneity was still high in
some analyses, possibly because of the high risk of method-
ological bias observed in most studies (mainly concerning the
sample selection), publication bias, and different methods used
to perform the visual inspection. Actually, differences in the
studies due to disparities in the threshold for a positive result
for either the diagnostic or reference standard method used
have been pointed out as common reasons of heterogeneity in

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015

Visual Inspection for Caries Detection 9
systematic reviews of diagnostic methods (Lijmer et al. 2002;
Dinnes et al. 2005).
Concerning the metaregressions performed in terms of the
reference standard methods, at the initial carious lesion thresh-
old, we observed a better performance in studies using opera-
tive intervention as the reference standard on occlusal surfaces
of permanent teeth, probably due to incorporation bias (Lijmer
et al. 1999). Usually, the reference standard method in clinical
studies of occlusal surfaces is operative intervention.
Consequently, this method could not be employed in the entire
sample, because teeth thought of as being sound would not be
opened by operative intervention to confirm that they were
sound, for ethical reasons.
The most important finding of this review concerns utiliza-
tion of validated visual scoring systems in the detection of cari-
ous lesions, which has been proposed as a way of improving
sensitivity and reliability (Braga et al. 2010). We found that
studies using well-established scoring systems showed higher
accuracy in detecting carious lesions in the proximal and
occlusal surfaces of primary and permanent teeth when com-
pared with studies that did not report the method used or that
used their own criteria. The utilization of a validated system
allows a more accurate diagnosis of caries lesions since it gives
practitioners a guideline to how to characterize the lesions that
they identify and it provides a rational shortcut associating
with these characteristics. This is the first robust study that has
proved empirically that utilization of a validated visual scoring
system improves the accuracy in detection of caries lesions.
In conclusion, visual inspection presents good accuracy in
the detection of carious lesions in primary and permanent
teeth, with a trend for higher specificity than sensitivity.
Importantly, the use of detailed and validated methods
improves the accuracy of visual inspection for caries.
Implications for Clinical Practice
Since visual inspection presented a good overall accuracy and
high value for specificity, we recommend its utilization in clini-
cal practice alone, with no need for an adjunct method. Regarding
the utilization of indices, we strongly encourage and recommend
the utilization of a well-established visual scoring system, since
it can significantly increase overall accuracy of the method.
Implications for Research
Most studies were performed in a laboratory, and this exerts a
significant influence on the overall specificity and sensitivity
values. If this type of study is needed, researchers should be
aware of their possible limitations, since these results cannot
be reliably extrapolated for clinical practice; as such, their use
should be limited.
Author Contributions
T. Gimenez, F.M. Mendes, contributed to conception, design, data
acquisition, analysis, and interpretation, drafted and critically
revised the manuscript; C. Piovesan, contributed to data acquisition,
Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.
© International & American Associations for Dental Research 2015
critically revised the manuscript; M.M. Braga, contributed to con-
ception, design, data acquisition, and interpretation, critically
revised the manuscript; D.P. Raggio, C. Deery, D.N. Ricketts, K.R.
Ekstrand, contributed to conception and data interpretation, criti-
cally revised the manuscript. All authors gave final approval and
agree to be accountable for all aspects of the work.
Acknowledgments
The study was supported by the Conselho Nacional de
Desenvolvimento Científico e Tecnológico, Fundação de Amparo
à Pesquisa do Estado de São Paulo (Process 2012/17888-1), and
CAPES. The authors wish also to thank the participants of the
Post-graduation in Pediatric Dentistry Seminar of FOUSP for the
critical comments put forth. The authors declare no potential con-
flicts of interest with respect to the authorship and/or publication
of this article.
References
Bader JD, Shugars DA, Bonito AJ. 2002. A systematic review of the perfor-
mance of methods for identifying carious lesions. J Public Health Dent.
62(4):201–213.
Baelum V, Hintze H, Wenzel A, Danielsen B, Nyvad B. 2012. Implications of
caries diagnostic strategies for clinical management decisions. Community
Dent Oral Epidemiol. 40(3):257–266.
Braga MM, Mendes FM, Ekstrand KR. 2010. Detection activity assessment
and diagnosis of dental caries lesions. Dent Clin North Am. 54(3):479–
493.
Braga MM, Moraes CC, Nakama RC, Leamari VM, Siqueira WL, Mendes
FM. 2009. In vitro performance of methods of approximal caries detection
in primary molars. Oral Surg Oral Med Oral Pathol Oral Radiol Endod.
108(4):e35–e41.
Deville WL, Bezemer PD, Bouter LM. 2000. Publications on diagnostic test
evaluation in family medicine journals: an optimal search strategy. J Clin
Epidemiol. 53(1):65–69.
Dinnes J, Deeks J, Kirby J, Roderick P. 2005. A methodological review of how
heterogeneity has been examined in systematic reviews of diagnostic test
accuracy. Health Technol Assess. 9(12):1–113.
Downer MC. 1975. Concurrent validity of an epidemiological diagnostic sys-
tem for caries with the histological appearance of extracted teeth as validat-
ing criterion. Caries Res. 9(3):231–246.
Downer MC. 1989. Validation of methods used in dental caries diagnosis. Int
Dent J. 39(4):241–246.
Ekstrand KR, Ricketts DN, Kidd EA. 1997. Reproducibility and accuracy of
three methods for assessment of demineralization depth of the occlusal sur-
face: an in vitro examination. Caries Res. 31(3):224–231.
Fyffe HE, Deery C, Nugent ZJ, Nuttall NM, Pitts NB. 2000. In vitro validity
of the Dundee Selectable Threshold Method for caries diagnosis (DSTM).
Community Dent Oral Epidemiol. 28(1):52–58.
Gimenez T, Braga MM, Raggio DP, Deery C, Ricketts DN, Mendes FM. 2013.
Fluorescence-based methods for detecting caries lesions: systematic review,
meta-analysis and sources of heterogeneity. PLoS One. 8(4):e60421.
Gimenez T, Piovesan C, Braga MM, Raggio DP, Deery C, Ricketts DN,
Ekstrand KR, Mendes FM. 2015. Clinical relevance of studies on accuracy
of visual inspection for detecting caries lesions: a systematic review. Caries
Res. 49(2):91–98.
Grimes DA, Schulz KF. 2002. An overview of clinical research: the lay of the
land. Lancet. 359(9300):57–61.
Ismail AI. 2004. Visual and visuo-tactile detection of dental caries. J Dent Res.
83 Spec No C:C56–C66.
Ismail AI, Sohn W, Tellez M, Amaya A, Sen A, Hasson H, Pitts NB. 2007.
The International Caries Detection and Assessment System (ICDAS):
an integrated system for measuring dental caries. Community Dent Oral
Epidemiol. 35(3):170–178.
Kuhnisch J, Goddon I, Berger S, Senkel H, Bucher K, Oehme T, Hickel R,
Heinrich-Weltzien R. 2009. Development, methodology and potential of
the new Universal Visual Scoring System (UniViSS) for caries detection
and diagnosis. Int J Environ Res Public Health. 6(9):2500–2509.
Juni P, Holenstein F, Sterne J, Bartlett C, Egger M. 2002. Direction and impact
of language bias in meta-analyses of controlled trials: empirical study. Int J
Epidemiol. 31(1):115–123.
10 Journal of Dental Research 

Lijmer JG, Bossuyt PM, Heisterkamp SH. 2002. Exploring sources of hetero-
geneity in systematic reviews of diagnostic tests. Stat Med. 21(11):1525–
1537.
Lijmer JG, Mol BW, Heisterkamp S, Bonsel GJ, Prins MH, van der Meulen JH,
Bossuyt PM. 1999. Empirical evidence of design-related bias in studies of
diagnostic tests. JAMA. 282(11):1061–1066.
Lussi A. 1993. Comparison of different methods for the diagnosis of fissure
caries without cavitation. Caries Res. 27(5):409–416.
Macaskill P, Gatsonis C, Deeks J, Harbord R, Takwoingi Y. 2010. Analysing
and presenting results. In: Deeks J, Bossuyt P, Gatsonis C, editors. Cochrane
handbook for systematic reviews of diagnostic test accuracy version 1.0.
London (UK): Cochrane Collaboration. p. 1–61.
Marthaler TM. 1966. A standardized system of recording dental conditions.
Helv Odontol Acta. 10(1):1–18.
Mendes FM, Novaes TF, Matos R, Bittar DG, Piovesan C, Gimenez T,
Imparato JC, Raggio DP, Braga MM. 2012. Radiographic and laser fluores-
cence methods have no benefits for detecting caries in primary teeth. Caries
Res. 46(6):536–543.
Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. 2009. Preferred
reporting items for systematic reviews and meta-analyses: the PRISMA
statement. PLoS Med. 6(7):e1000097.
Moher D, Pham B, Klassen TP, Schulz KF, Berlin JA, Jadad AR, Liberati A.
2000. What contributions do languages other than English make on the
results of meta-analyses? J Clin Epidemiol. 53(9):964–972.
Nelemans PJ, Leiner T, de Vet HC, van Engelshoven JM. 2000. Peripheral arte-
rial disease: meta-analysis of the diagnostic performance of MR angiogra-
phy. Radiology. 217(1):105–114.
Nytun RB, Raadal M, Espelid I. 1992. Diagnosis of dentin involvement in
occlusal caries based on visual and radiographic examination of the teeth.
Scand J Dent Res. 100(3):144–148.
Nyvad B, Machiulskiene V, Baelum V. 1999. Reliability of a new caries diag-
nostic system differentiating between active and inactive caries lesions.
Caries Res. 33(4):252–260.
Pitts N. 2004. “ICDAS”: an international system for caries detection and assess-
ment being developed to facilitate caries epidemiology, research and appro-
priate clinical management. Community Dent Health. 21(3):193–198.
Pitts NB, Evans DJ, Pine CM. 1997. British Association for the Study of
Community Dentistry (BASCD) diagnostic criteria for caries prevalence
surveys: 1996/97. Community Dent Health. 14(Suppl 1):6–9.
Sterne JA, Egger M, Moher D. 2011. Addressing reporting biases. In: Higgins
JP, Green S, editors. Cochrane handbook for systematic reviews of inter-
vations version 5.1.0. London (UK): Cochrane Collaboration. Chapter 10.
Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB,
Leeflang MM, Sterne JA, Bossuyt PM. 2011. QUADAS-2: A revised tool
for the quality assessment of diagnostic accuracy studies. Ann Intern Med.
155(8):529–536.
World Health Organization. 1997. Oral health surveys, basic methods. 4th ed.
Geneva (Switzerland): World Health Organization.

Downloaded from jdr.sagepub.com at NORTH DAKOTA STATE UNIV LIB on May 25, 2015 For personal use only. No other uses without permission.

© International & American Associations for Dental Research 2015