Abstract
Yardsticks have been developed to measure dental arch relations in cleft lip and palate (CLP) patients as diagnostic proxies for the underlying skeletal relationship. Travelling with plaster casts to compare results between CLP centres is inefficient so the aim of this study was to investigate the reliability of using digital models or photographs of dental casts instead of plaster casts for rating dental arch relationships in children with complete bilateral cleft lip and palate (CBCLP). Dental casts of children with CBCLP ( n = 20) were included. Plaster casts, digital models and photographs of the plaster casts were available for all the children at 6, 9, and 12 years of age. All three record formats were scored using the bilateral cleft lip and palate (BCLP) yardstick by four observers in random order. No significant differences were found for the BCLP yardstick scores among the three formats. The interobserver weighted kappa scores were between 0.672 and 0.934. Comparison between the formats per observer resulted in weighted kappa scores between 0.692 and 0.885. It is concluded that digital models and photographs of dental casts can be used for rating dental arch relationships in patients with CBCLP. These formats are a reliable alternative for BCLP yardstick assessments on conventional plaster casts.
A multidisciplinary treatment approach is required in children with bilateral cleft lip and palate (BCLP). Such a treatment involves multiple surgeries and orthodontic treatment until facial growth and dental development have ceased. A reliable method for evaluating and comparing treatment outcomes, in particular dental arch relationships, is essential. For patients with a unilateral cleft lip and palate (UCLP) several scoring methods have been described . Two of these methods have been modified for patients with a complete bilateral cleft lip and palate (CBCLP) .
The modified H uddart /B odenham scoring system grades dental outcome by considering the bucco-palatal relationships in terms of frequency and severity of cross-bites in the anterior and buccal segments in order to evaluate maxillary constriction . The BCLP yardstick grades the dental arch relationship as a proxy for facial growth. This scoring system has the advantage over the previous one that it takes greater account of the skeletal component. It assesses dental relationship in terms of antero-posterior, transverse and vertical discrepancies. The sagittal dental base relationship is considered the most important feature and dental compensation is taken into account .
The BCLP yardstick appears to be a good tool to compare multiple cleft centres in their treatment results . It requires travelling with plaster casts to one place to perform comparative studies with the inconvenience of transport, expense and the possible damage of the models . For the Goslon yardstick, a scoring system for UCLP patients, several formats to replace the fragile plaster models have been tested . L iao et al. proved intraoral photographs to be a viable alternative to the application of the Goslon yardstick on dental casts . N ollet et al. described photographs of dental casts as a good and reliable alternative and suggest the use of digital models . For the BCLP yardstick an appropriate approach would be to substitute the plaster models by photographs or even digital models as an alternative medium. The purpose of this study was to investigate the reliability of using digital models or photographs of dental casts as an alternative to plaster casts for rating dental arch relationships in children with CBCLP.
Material and methods
Dental casts of 26 patients (20 in the experimental sample, 6 for calibration) with CBCLP were used. Patients with Simonart’s bands or syndromes were excluded. Three sets of dental casts were available for all patients at approximately the age of 6 (range 5.0–6.6), 9 (range 7.11–10.1), and 12 years (range 10.10–13.8).
Dental arch relationships were rated with the BCLP yardstick on a 5-point-scale. This is a rating system developed to grade dental arch relationships in the early, late mixed and early permanent dentition into five categories in terms of antero-posterior, transverse and vertical discrepancies in patients with CBCLP .
A score of 1 or 2 indicates Class II/1 and Class I dental relationship, a score of 3 means a less favourable antero-posterior relationship, an end-to-end situation, whilst patients with scores of 4 or 5 will probably require an osteotomy to correct the antero-posterior relationship . A set of 33 reference models representing the five categories of the yardstick is available to assist reproducible rating .
For grading with the BCLP yardstick on photographs of the dental casts, digital images of the plaster casts of the 26 patients at three different ages were obtained with a Nikon (Melville, NY, USA) D1 camera and an AF Micro Nikkon (Melville, NY, USA) 60 mm/1:2.8 D lens. The lens–object distance was about 30 cm. For each set of models, a frontal view, views from the right and left side and occlusal views of the upper and lower arch were made with a black background. The maximum time needed for photographing one set of models was 5 min. The pictures were loaded into PowerPoint (Microsoft Corp., Redmond, WA, USA). Each slide contained five views of a set of dental casts of one patient together with the identifying case number and the age ( Fig. 1 ). The photographs were displayed and rated on a laptop.
For the grading with the BCLP yardstick on digital models, all plaster casts were digitized (Orthoproof, Doorn, The Netherlands) according to a standardized procedure. The digital models were displayed on a laptop using the Digimodel ® programme ( Fig. 2 ) (Orthoproof, Doorn, The Netherlands).
Four calibrated observers carried out the ratings (CK, AK, GS, WS). The measurements were done in three rounds over 2 days. In every round, each observer rated 60 models (20 digital models, 20 cast models and 20 photographs, all in randomized order). After the first round every observer had seen all 20 patients at all three ages but not in all formats. On the first day, calibration of the observers was undertaken on six patients who were not included in the experimental sample.
Statistics
In analysing observer performance, the BCLP yardstick can be treated as a categorical scale, which implies the use of the weighted kappa statistic. Since the BCLP yardstick is an ordinal scale, its scores can be treated as nominal scores as well . In that case the reliability coefficient, the duplicate measurement error and the systematic difference between paired observations are the evaluation parameters. The authors used both approaches. Models from all three age groups were analysed simultaneously. For each format all combinations of interobserver performance were analysed and between the formats all intraobserver comparisons were made. Additionally per format the mean score of the observer was calculated, in order to allow comparisons between the formats irrespective of the observer. This can only be done from the numerical interpretation of the BCLP yardstick. For the categorical interpretation the weighted kappa was applied to the accumulated intra-comparisons of the four observers .
Respective reproducibility for the three formats was analysed. To this end ANOVA was used for three outcomes: weighted kappa, reliability coefficient and duplicate measurement error.
Results
Interobserver performance for three formats
Table 1 shows the reliability coefficients ranging from 0.817 to 0.970. Six combinations of observers per method were made. In the 18 comparisons, 7 showed a statistically significant difference between the observers. These were seen equally in the three methods (three in plaster casts, two in digital models, two in pictures of the models). The biggest difference found was 0.233 BCLP yardstick points.
Observer | Reliability | Mean diff. | 95% CI of mean diff. | P -value | DME | Weighted kappa | |
---|---|---|---|---|---|---|---|
Plaster cast vs. plaster cast | A–B | 0.861 | −0.067 | [−0.223…0.090] | 0.398 | 0.429 | 0.742 |
A–C | 0.914 | −0.017 | [−0.138…0.105] | 0.784 | 0.332 | 0.838 | |
A–D | 0.919 | 0.150 | [0.035…0.265] | 0.011 | 0.314 | 0.830 | |
B–C | 0.868 | 0.050 | [−0.104…0.204] | 0.517 | 0.420 | 0.757 | |
B–D | 0.866 | 0.217 | [0.066…0.368] | 0.006 | 0.414 | 0.716 | |
C–D | 0.945 | 0.167 | [0.070…0.264] | 0.001 | 0.266 | 0.846 | |
Digital model vs. digital model | A–B | 0.914 | 0.083 | [−0.036…0.203] | 0.167 | 0.327 | 0.838 |
A–C | 0.836 | 0.167 | [0.008…0.326] | 0.040 | 0.435 | 0.722 | |
A–D | 0.842 | 0.200 | [0.044…0.356] | 0.013 | 0.428 | 0.723 | |
B–C | 0.840 | 0.083 | [−0.076…0.243] | 0.301 | 0.437 | 0.768 | |
B–D | 0.829 | 0.117 | [−0.049…0.282] | 0.163 | 0.453 | 0.737 | |
C–D | 0.970 | 0.033 | [−0.033…0.100] | 0.321 | 0.183 | 0.934 | |
Picture of plaster cast vs. picture of plaster cast | A–B | 0.884 | 0.183 | [0.053…0.314] | 0.007 | 0.356 | 0.733 |
A–C | 0.887 | 0.100 | [−0.023…0.223] | 0.109 | 0.337 | 0.802 | |
A–D | 0.901 | 0.233 | [0.113…0.353] | 0.000 | 0.328 | 0.780 | |
B–C | 0.853 | −0.083 | [−0.228…0.062] | 0.255 | 0.397 | 0.723 | |
B–D | 0.858 | 0.050 | [−0.096…0.196] | 0.496 | 0.400 | 0.755 | |
C–D | 0.817 | 0.133 | [−0.028…0.294] | 0.103 | 0.441 | 0.672 |
The duplicate measurement errors varied from 0.183 to 0.453 BCLP yardstick points, and the weighted kappa ranged between 0.672 and 0.934.
Comparison between the three formats per observer
Table 2 shows the reliability coefficients ranging from 0.833 to 0.949. Three formats were compared for four observers. Amongst the 12 comparisons, only 3 showed a statistically significant difference. The biggest mean difference found was 0.250 BCLP yardstick points.
Observer | Reliability | Mean diff. | 95% CI of mean diff. | P -value | DME | Weighted kappa | |
---|---|---|---|---|---|---|---|
Digital mode vs. plaster cast | A–A | 0.877 | 0.000 | [−0.143…0.143] | 1.000 | 0.391 | 0.793 |
B–B | 0.887 | −0.150 | [−0.291…−0.009] | 0.038 | 0.387 | 0.758 | |
C–C | 0.911 | −0.183 | [−0.305…−0.062] | 0.004 | 0.332 | 0.770 | |
D–D | 0.946 | −0.050 | [−0.138…0.038] | 0.260 | 0.241 | 0.885 | |
Plaster cast vs. picture of plaster cast | A–A | 0.842 | 0.000 | [−0.158…0.158] | 1.000 | 0.432 | 0.724 |
B–B | 0.833 | 0.250 | [0.081…0.419] | 0.004 | 0.463 | 0.692 | |
C–C | 0.842 | 0.117 | [−0.042…0.275] | 0.146 | 0.434 | 0.733 | |
D–D | 0.949 | 0.083 | [−0.003…0.170] | 0.058 | 0.236 | 0.885 | |
Digital model vs. picture of plaster cast | A–A | 0.881 | 0.000 | [−0.135…0.135] | 1.000 | 0.368 | 0.815 |
B–B | 0.836 | 0.100 | [−0.063…0.263] | 0.224 | 0.445 | 0.754 | |
C–C | 0.839 | −0.067 | [−0.216…0.083] | 0.376 | 0.409 | 0.735 | |
D–D | 0.940 | 0.033 | [−0.061…0.128] | 0.484 | 0.259 | 0.868 |