Professional Documents
Culture Documents
Ekstrand 2018 PDF
Ekstrand 2018 PDF
Ekstrand 2018 PDF
Mariana M. Braga b
Chawla et al. [2012] 0.55 (0.35, 0.75) Neuhaus et al. [2015a] 0.41 (0.20, 0.62)
Freitas et al. [2016] 0.75 (0.67, 0.83) Achilleos et al. [(2013] 0.74 (0.23, 1.25)
Braga et al. [2009b] 0.95 (0.80, 1.10) Ozturk and Sinanoglu [2015] 0.71 (0.51, 0.91)
Bussaneli et al. [2015] 0.75 (0.40, 1.10) Neuhaus et al. [2015b]* 0.57 (0.33, 0.81)
Noaves et al. [2012] 0.76 (0.55, 0.96) Neuhaus et al. [2015b]*** 0.50 (0.26, 0.74)
Neuhaus et al. [2011] 0.73 (0.21, 1.25) Jallad et al. [2015] 0.88 (0.80, 0.96)
Ekstrand et al. [2011] 0.96 (0.83, 1.09) Diniz et al. [2011] 0.89 (0.86, 0.92)
Souza et al. [2013] 0.92 (0.72, 1.12) Gomez et al. [2013] 0.85 (0.56, 1.14)
Cotta et al. [2015] 0.84 (0.57, 1.11)
DOI: 10.1159/000486429
RE model 0.81 (0.71, 0.92) Rodrigues et al. [2008] 0.61 (0.29, 0.93)
Braga
Ekstrand/Gimenez/Ferreira/Mendes/
(For legend see next page.)
Downloaded by:
Univ. of California Santa Barbara
128.111.121.42 - 3/9/2018 3:15:09 PM
Interexaminer reproducibility
Chawla et al. [2012] 0.41 (0.35, 0.47) Neuhaus et al. [2015a] 0.64 (0.42, 0.86)
Teo et al. [2014] 0.84 (0.61, 1.07) Mitropoulos et al. [2010] 0.51 (0.36, 0.66)
Braga et al. [2010] 0.91 (0.69, 1.13) Parviainen et al. [2013] 0.47 (0.02, 0.92)
Fig. 1. Forest plots for meta-analysis and heterogeneity analysis of intra- and inter-examiner reproducibility when ICDAS was used for assessing caries lesion severity.
DOI: 10.1159/000486429
Caries Res 2018;52:406–419
IRR, intra- or inter-rater reproducibility. Examiners: * experienced; ** intermediate (+training); *** novice. Surfaces: # approximal; ## smooth; ### occlusal.
409
Downloaded by:
Univ. of California Santa Barbara
128.111.121.42 - 3/9/2018 3:15:09 PM
Study selection was performed independently by 2 reviewers [CRD, 2009]. The trim and fill method was used to adjust the
(T.G. and M.M.B.). Disagreements were resolved by discussion pooled estimates, when necessary (R package meta, version 4.8-2)
with a third researcher (K.R.E.). For studies with the same data set, [Schwarzer, 2007].
only the study considered the most complete was included in this
review. If the same study evaluated the accuracy of different sub- Systematic Review: Activity Assessment Associated with the
groups, the data were recorded as different studies. ICDAS
Information Sources, Study Selection, and Eligibility Criteria
Data Collection Process For the second review, the focus of attention was to collect in-
We searched papers for intra- and inter-examiner reproduc- formation about the systems associated with the ICDAS for activ-
ibility data of the ICDAS including quantitative intra- or inter- ity assessment of coronal caries lesions and, when possible, to ex-
rater reliability coefficients (intra-RR or inter-RR). When more press their performance in terms of reproducibility and accuracy.
than 1 examiner was involved, the mean values of the coefficients Following the question: “How has the activity assessment been
were used. If the study presented any estimation of variance, they performed when the ICDAS is used?” another search was per-
were also registered. If not, the standard error was estimated using formed in PubMed until January 2017 to summarize the evidence
a graphical model [Hanley, 1987]. about activity assessment using the ICDAS (online suppl. Fig. S1).
We also collected the absolute numbers of true positives, false We screened studies that combined any type of activity assess-
positive, true negatives, and false negatives or data to calculate ment and the visual examination performed by the ICDAS, how-
these figures. If the study had evaluated the performance of the ever exclusively on coronal caries. Articles that did not report data
method with more than 1 examiner, we used the values of the first about performance of the activity assessment or only reported in
examiner. If examiners with different levels of training participat- vitro experiments were excluded. The same strategy described in the
ed, data from the first examiner per category of training were col- previous section was used for duplicates. The selection of studies
lected and each category was considered as an independent set of was also performed by 2 independent reviewers (F.R.F. and M.M.B.)
data. Data about the correlation between the ICDAS scores and the and the same strategies were used for solving disagreements.
histological validation were also collected.
Data Collection Process and Synthesis of Results
Summary Measures and Synthesis of Results We summarized the studies that used the activity assessment
Subgroups concerning the type of teeth and study setting were associated with the ICDAS and their purposes. When aiming to
considered for all analyses. For the meta-analyses of intra-RR and validate the method, we also registered how it was performed. For
inter-RR, random-effects models were fitted using the R package longitudinal studies, findings regarding caries progression related
metafor (GNU General Public License version 2) [Viechtbauer, to the activity status of caries lesions were collected. When report-
2010]. Separate analyses were conducted for intra-RR and inter-RR ing data about performance, we carried out data collection and
data. When more than 1 IRR value was collected from the article as analyses similarly to in the first section.
reproducibility measurements, we preferred the following order of For the accuracy of activity assessment systems, sound surfaces
IRR coefficients to be used in these analyses: weighted Kappa, un- were not included in these analyses. Pooled results were presented
weighted Kappa, and intraclass correlation coefficient. Coefficients for all merged systems since they evaluated similar characteristics.
were pooled according to the methodology of a previously pub-
lished study [Bornmann et al., 2010]. Concerning the correlation Risk of Bias of Individual Studies
coefficients between the ICDAS and histology, pooled coefficients Two reviewers (M.M.B. and T.G.) were responsible for risk of
and their 95% confidence intervals are summarized in forest plots. bias assessment in the reviews reported above. The QUADAS-2
For accuracy, statistical analyses were performed for the sub- was used to assess the potential risk of bias of included test accu-
groups above and also the thresholds (D1 or D3). At D1 threshold, racy studies in both reviews. According to 4 domains (sample se-
all lesions were considered as carious. At D3, only more advanced lection, index test, reference standard, or timing and flow), the
caries lesions were included. Thus, surfaces considered as carious for studies were classified as low, high, or unclear risk of bias.
these analyses were those lesions into dentine when lesion depth was For longitudinal studies, the Newcastle-Ottawa quality assess-
assessed, or cavitated lesions when surface integrity was considered. ment scale was used to classify the risk of bias in 3 different do-
Statistical pooling of sensitivity, specificity, positive likelihood mains: selection, comparability, and outcome. This system identi-
(sensitivity/1 – specificity) and negative likelihood (1 – sensitivity/ fies the risk of bias using stars. If most of the stars were awarded
specificity), and diagnostic odds ratio [DOR = (TP/FP)/(FN/TN)] for the domain, the study was classified as having a low risk of bias
was carried out using the DERSimonian Laid method (random ef- considering this aspect.
fects meta-analyses model). We summarized these analyses in re-
ceiver operating characteristics curves (SROC) and tested the het-
erogeneity among the included studies using the Cochran Q and I2
tests. Sensitivity analyses were performed to check the influence of Results
studies with approximal caries lesions on the pooled estimates. All
these analyses were performed in MetaDisc 1.4 software (Unidad de Systematic Review: The Accuracy of the ICDAS in
Bioestadistica Clinica del Hospital Ramón y Caja, Madrid, Spain). Detecting Coronal Caries
Publication bias was assessed by checking the asymmetry of
funnel plots visually and using regression analysis performed in R Synthesis of Results
package metafor. Alternatively, for accuracy, we used funnel plots A total of 54 studies from 46 papers fulfilled the re-
of natural logarithm (ln) DOR versus 1/√effective sample size quirements for enrolment (online suppl. Fig. S2). Most of
128.111.121.42 - 3/9/2018 3:15:09 PM
Univ. of California Santa Barbara
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1 – specificity 1 – specificity
Primary teeth
Permanent teeth
Clinical studies
SROC curve SROC curve
1.0
n studies = 6 n studies = 4
0.9
AUC(SE) = 0.7908 AUC(SE) = 0.7820
0.8 (0.0297) (0.0420)
Pooled Sen = 0.553 Pooled Sen = 0.563
0.7 (0.521–0.585) (0.523–0.602)
Pooled Spe = 0.861 Pooled Spe = 0.966
(0.831–0.888) (0.958–0.973)
0.6
(REM) pooled LR+ = 2.647 (REM) pooled LR+ = 2.889
Sensitivity
(2.103–3.332) (0.594–14.060)
0.5
(REM) pooled LR– = 0.407 (REM) pooled LR– = 0.438
(0.261–0.635) (0.242–0.792)
0.4 (REM) pooled DOR = (REM) pooled DOR =
6.118 (3.988–9.384) 7.522 (1.519–37.255)
0.3 Heterogeneity Q (SE) = Heterogeneity Q (SE) =
8.33 (d.f. = 5) p = 0.139 20.55 (d.f. = 3) p < 0.001
Inconsistency I2 = 40.0% Inconsistency I2 = 85.4%
0.2
Estimate of between- Estimate of between-
study variance τ = 0.1045 study variance τ = 2.1578
0.1
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1 – specificity 1 – specificity
Fig. 2. Summary ROC diagrams for meta-analysis and heterogeneity analysis of ICDAS accuracy on assessing
any caries lesions at the D1 threshold – number of studies, area under the ROC curves, pooled sensitivity (Sen),
specificity (Spe), DOR, positive and negative likelihood ratios (LR), results for the Cochran Q test for heteroge-
neity, I2 and τ coefficient variances between studies.
(4.423–12.999) (3.862–9.188)
0.5 (REM) pooled LR– =0.339 (REM) pooled LR– = 0.253
(0.235–0.488) (0.138–0.464)
0.4 (REM) pooled DOR = (REM) pooled DOR =
33.866 (13.165–87.115) 26.937 (13.701–52.959)
0.3 Heterogeneity Q (SE) = Heterogeneity Q (SE) =
92.31 (d.f. = 15) p < 0.001 98.27 (d.f. = 21), p < 0.001
Inconsistency I2 = 83.7% Inconsistency I2 = 78.6%
0.2 Estimate of between- Estimate of between-study
study variance τ2 = 2.9472 variance τ2 = 1.8415
0.1
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Permanent teeth
1 – specificity 1 – specificity
Primary teeth
Clinical studies
SROC curve SROC curve
1.0
n studies = 6
0.9 n studies = 5*
AUC(SE) = 0.9246
0.8 (0.0384) AUC(SE) = 0.6763 (0.0792)
Pooled Sen = 0.598 Pooled Sen = 0.545 (0.475–
(0.504–0.688) 0.614)
0.7 Pooled Spe = 0.971 Pooled Spe = 0.972
(0.962–0.978) (0.965–0.977)
0.6 (REM) pooled (REM) pooled LR+ = 6.419
LR+ = 11.478 (3,763– (1.744–23.619)
Sensitivity
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1 – specificity 1 – specificity
Fig. 3. Summary ROC diagrams for meta-analysis and heterogeneity analysis of ICDAS accuracy on assessing
more advanced caries lesions (cavitated caries lesions or those lesions histologically into dentine/at the D3 thresh-
old – number of studies, area under the ROC curves, pooled sensitivity (Sen), specificity (Spe), DOR, positive and
negative likelihood ratios (LR), results for the Cochran Q test for heterogeneity, I2 and τ coefficient variances
between studies. * One study was omitted in the pooled analysis because of a false-positive value = 0, Spe = 1.
The pooled Spearman correlation between the ICDAS trast, the pooled sensitivities values were lower in the clin-
scores and histology was 0.78 for primary teeth (95% CI ical studies (0.55–0.60), but higher in the laboratory stud-
0.73–0.84; n = 5) and 0.68 for the permanent teeth (95% ies (>0.70). Pooled likelihood ratios (LR+) were above the
CI 0.58–0.78; n = 12). As very few studies were performed value 1 (2.65–11.48) and LR data ranged from 0.15 to
in vivo, we presented the global pooled results. 0.51. The DOR values were very high (6.1–59.6; Fig. 2, 3).
The areas under the ROC curves were all substantial Values of SROC, sensitivity and specificity were simi-
(≥0.75), apart from permanent teeth in clinical studies, lar between D1 and D3 levels for laboratorial studies
where the area under the ROC curve was 0.67 (Fig. 2, 3). (Fig. 2, 3). At the D1 threshold, slightly higher SROC val-
The pooled specificities were also all high (>0.70). In con- ues were found for laboratorial conditions. We also no-
128.111.121.42 - 3/9/2018 3:15:09 PM
Univ. of California Santa Barbara
0.8
Pooled sensitivity = 0.70 (0.63–0.77)
0.7 χ2 = 20.66: d.f. = 3 (p = 0.0001)
Inconsistency I2 = 85.5%
0 0.2 0.4 0.6 0.8 1.0
0.6
Sensitivity
Sensitivity
0.5
0.4
0.3
Specificity (95% CI)
0.2 Freitas et al. [2016] 0.64 (0.45–0.80)
Cotta et al. [2015] 0.40 (0.24–0.58)
Braga et al. [2010] 0.17 (0.02–0.48)
0.1
Ekstrand et al. [2007] 0.81 (0.64–0.93)
Fig. 4. Meta-analysis of ICDAS+LAA accuracy on assessing caries lesion activity (left: summary ROC diagram;
right: forest plots for meta-analysis using sensitivity and specificity of the method).
ticed a trend of lower pooled sensitivities for clinical stud- ing primary teeth. At the D3 threshold, for the same set-
ies; however, these were associated with higher specificity ting, only sensitivity was increased (0.86; 95% CI 0.76–
values (Fig. 2). For clinical studies, a slightly lower pooled 0.94).
sensitivity was observed at the D3 level. For more ad- Apart from clinical studies conducted with primary
vanced caries lesions, a very high pooled specificity teeth, moderate to high levels of heterogeneity (I2) were
(>0.90) was observed for primary teeth, while a very low observed among the studies (Fig. 4, 5).
pooled specificity was found for permanent teeth (Fig. 3).
A wide variety of results was observed, especially among Systematic Review: Activity Assessment Associated
studies using permanent teeth (Fig. 3). with the ICDAS
Looking to the SROC, we observed some very discrep- Synthesis of Results
ant points, which represent some of the studies per- Thirty-six studies were retrieved in this search (online
formed with approximal lesions (Fig. 2, 3). When exclud- suppl. Fig. S3). Out of 11 included studies, 9 were cross-
ing the studies with approximal surfaces from the sample sectional and 2 were longitudinal studies (online suppl.
(sensitivity analysis), the DOR was increased by 20% for Table S2). Five studies validated a system to be used in
laboratory studies concerning permanent teeth at the D1 caries activity assessment in conjunction with the ICDAS.
threshold. For clinical studies, the exclusion of this type One additional study showed data compatible with a pre-
of surface resulted in a significantly higher sensitivity dictive validation of activity assessment of caries lesions
(0.88; 95% CI 0.81–0.93), AUC (0.90; 95% CI 0.83–0.97), [Ferreira Zandona et al., 2012]. Four studies were inter-
and DOR (14.860; 95% CI 6.756–32.670) at D1 concern- ested in the relationship between the ICDAS and activity
128.111.121.42 - 3/9/2018 3:15:09 PM
Univ. of California Santa Barbara
References
Achilleos EE, Rahiotis C, Kakaboura A, Vou- Bornmann L, Mutz R, Daniel HD: A reliability- Braga MM, Morais CC, Nakama RC, Leamari
giouklakis G: Evaluation of a new fluores- generalization study of journal peer reviews: a VM, Siqueira WL, Mendes FM: In vitro per-
cence-based device in the detection of incipi- multilevel meta-analysis of inter-rater reli- formance of methods of approximal caries de-
ent occlusal caries lesions. Lasers Med Sci ability and its determinants. PLoS One 2010; tection in primary molars. Oral Surg Oral
2013;28:193–201. 5:e14331. Med Oral Pathol Oral Radiol Endod 2009b;
Agustsdottir H, Gudmundsdottir H, Eggertsson Braga MM, Ekstrand KR, Martignon S, Imparato 108:e35–41.
H, Jonsson SH, Gudlaugsson JO, Saemunds- JC, Ricketts DN, Mendes FM: Clinical perfor- Bussaneli DG, Boldieri T, Diniz MB, Rivera LM,
son SR, Eliasson ST, Arnadottir IB, Holbrook mance of two visual scoring systems in detect- Santos-Pinto L, Cordeiro Rde C: Influence of
WP: Caries prevalence of permanent teeth: a ing and assessing activity status of occlusal professional experience on detection and
national survey of children in Iceland using caries in primary teeth. Caries Res 2010; 44: treatment decision of occlusal caries lesions in
ICDAS. Community Dent Oral Epidemiol 300–308. primary teeth. Int J Paediatr Dent 2015; 25:
2010;38:299–309. Braga MM, Mendes FM, Martignon S, Ricketts 418–427.
Bader JD, Shugars DA, Bonito AJ: Systematic re- DN, Ekstrand KR: In vitro comparison of Castilho LS, Cotta FV, Bueno AC, Moreira AN,
views of selected dental caries diagnostic and Nyvad's System and ICDAS-II with lesion ac- Ferreira EF, Magalhaes CS: Validation of
management methods. J Dent Educ 2001; 65: tivity assessment for evaluation of severity DIAGNOdent laser fluorescence and the In-
960–968. and activity of occlusal caries lesions in pri- ternational Caries Detection and Assessment
Baelum V, Heidmann J, Nyvad B: Dental caries mary teeth. Caries Res 2009a;43:405–412. System (ICDAS) in diagnosis of occlusal car-
paradigms in diagnosis and diagnostic re- ies in permanent teeth: an in vivo study. Eur J
search. Eur J Oral Sci 2006;114:263–277. Oral Sci 2016;124:188–194.
128.111.121.42 - 3/9/2018 3:15:09 PM
Univ. of California Santa Barbara