Journal of Orthodontics, Vol.

34, 2007, 234–242

SCIENTIFIC Tooth-size discrepancy and Bolton’s

SECTION
ratios: the reproducibility and speed of
two methods of measurement
S. A. Othman and N. W. Harradine
Bristol Dental Hospital, Bristol, UK

Objective: To determine and compare the reproducibility and speed of two methods of performing Bolton’s tooth-size analysis.
Design: Analysis of randomly selected clinical sample.
Setting: Bristol Dental Hospital, University of Bristol, United Kingdom.
Materials and methods: Pre-treatment study casts of 150 patients were selected randomly from 1100 consecutively treated
Caucasian orthodontic patients. Bolton tooth-size discrepancies and ratios were measured using two methods; one method
employed entirely manual measurement and the Odontorule slide rule, while the other employed digital callipers and the
HATS analysis software. Twenty study casts were measured twice, a week apart with both methods. Another three
investigators also measured 20 study casts twice with the HATS analysis.
Results: There were small or no systematic errors within or between these two methods. A very significant difference was
evident for mean time measurements between the two methods (mean time for HATS was 3.5 minutes and for Odontorule was
8.9 minutes). There was relatively high error variance of both methods of measurement as a percentage of the total variance.
Conclusions: On-line electronic measurement was found to be more rapid than the manual method used. Both methods
demonstrate relatively high random error and this has important consequences for the clinical use of Bolton’s ratios.
Key words: Bolton’s ratios, tooth-size discrepancy, reproducibility, methods of measurement

Received 18th August 2006; accepted 7th July 2007

Introduction space analysis is that the computer can quickly provide a

subsequent tooth-size analysis.
Many studies have investigated tooth-size discrepancy Previous studies of Bolton’s ratios have included very
(TSD).1–4 The best-known study of TSD in relation to incomplete investigation or reporting of the reproduci-
treatment of malocclusion was published by Bolton5 in bility of their measurements.7–9 Othman and
1958 in which he evaluated 55 cases with excellent Harradine10 in their review paper on TSD concluded
occlusions. Bolton developed two ratios for estimating that reproducibility of measurement of TSD has been
TSD by measuring the summed mesiodistal widths of poorly investigated. Some well-known studies did not
the mandibular to the maxillary anterior teeth (anterior report the measurement error at all.11 Crosby and
ratio) and the total width of all lower to upper teeth Alexander12 did not mention which statistical test was
from first molar to first molar (overall ratio). used or what values were tested for the measurement
The traditional methods of measuring mesiodistal error or whether systematic error, random error or both
widths of teeth on dental casts have been either with types of error were evaluated. Araujo and Souki13 found
needle-pointed dividers or a Boley gauge (Vernier that there were no significant differences between two
callipers). Recent technological advances have allowed sets of measurements with digital callipers, but there was
the introduction of digital callipers, which can be linked no measurement of random error. Also, their analysis
to computers for quick calculation of the anterior and was confined to Bolton’s ratios, and did not include data
posterior ratios. Alternatively, digitized or scanned on the required corrections in millimetres, which is
images of study casts can be measured on-screen to the therapeutically relevant measurement. Conversely,
assist with diagnosis and treatment planning. Proffit6 Bernabé et al.14 only evaluated random error not
stated that one advantage of digital measurement for systematic error, and they too only examined the ratio

Address for correspondence: Mr N. W. T. Harradine, Department

of Child Dental Health, University of Bristol Dental School, Lower
Maudlin Street, Bristol BS1 2LY, UK.
Email: [email protected]
# 2007 British Orthodontic Society DOI 10.1179/146531207225022302
 JO December 2007 Scientific Section Tooth-size discrepancy and Bolton’s ratios 235

values but not the correction values in millimetres. If a points. Contact points were defined at the points on the
measurement is to be used to determine a therapeutic proximal surfaces, as observed or estimated as those
intervention, then it is important to know the measure- which should be touching when the teeth were perfectly
ment errors in relation to the planned dimensions of that aligned.
intervention. Houston15 wrote that if any study using Method 1: this employed the Odontorule for analysis
measurements is to be of value, it is imperative that such of the maxillary–mandibular tooth size relationship.
error analysis be undertaken and reported, and the very This employs a rotating wheel, which is in-effect a
discipline of undertaking an error analysis should also circular slide rule, and was developed by Dr David C.
improve the quality of results. Hamilton and Dr Charles W. Patton based upon studies
The aims and objectives of the present study were to by Dr Wayne A. Bolton as an aid to measurement which
determine and compare the reproducibility and speed of would be faster and more convenient than looking up
two methods of performing Bolton’s tooth-size analysis. tables of figures. The mesiodistal tooth sizes were
The two methods were a manual measurement with measured in millimetres manually to the nearest
the Odontorule (Dental Corporation of America, West 0.5 mm with Helios sliding callipers. The sum of the
Chester, PA, USA) and a computerized method – the total maxillary and mandibular teeth (6–6) and sum of
Hamilton Arch Tooth System (HATS) (GAC the anterior maxillary and mandibular teeth (3–3) were
International, Central Islip, NY, USA). The null calculated using a calculator. The total and anterior
hypothesis was that there would be no difference ratios were determined by Bolton’s formulas.5 The
between the methods with regard to rapidity or amounts of correction in the maxillary and mandibular
reproducibility. The central hypothesis was that at least arches for the total ratio and anterior ratio are obtained
one method would be sufficiently rapid and sufficiently on the rotating wheel. Each analysis was timed by a
reproducible to be a robust and practicable method of stopwatch from the first measurement to the final
measurement for an individual clinical case. computation.
Method 2: this employed the HATS software which is
Materials and methods available from GAC. All the study casts were measured
to the nearest 0.01 mm with digital callipers (PRO-
Pre-treatment study casts of 150 patients were used in MAX Digital Callipers, Fred V. Fowler Co., Inc.,
this study. The study model numbers of 1100 patients Newton, MA, USA) connected to a computer. The
treated consecutively in a teaching hospital from 1999 to HATS software calculates the Bolton’s ratios and also
2002 were obtained from the laboratory database and a recommends the tooth size correction in either arch to
computer-generated list of random numbers was used to achieve Bolton’s average ratio for an ideal occlusion.
select the sample from this consecutive series. If a case The entire procedure was timed from initial measure-
was discarded because it did not meet the selection ment to availability of the calculated results and the
criteria, the next consecutive eligible case was included. results were then printed. Neither the method of
This sample therefore included a random selection of measurement nor the timing of the measurements was
different malocclusions representative of an orthodontic amenable to blinding of the assessor at recording.
treatment population. The Chairman of the Local
Research Ethics Committee confirmed that ethical Assessment of reproducibility
approval for measuring study casts was not required.
The following selection criteria were used: The principal intra-examiner reproducibility procedure
consisted of the primary investigator (SAO) measuring
N good quality pre-treatment models;
20 sets of study models randomly selected from the
N a fully erupted permanent dentition from first molar
larger group of 150 patients. These patients were
to first molar;
selected from the 150 by means of a computer-generated
N Caucasian ethnicity.
random number list. A sample of 20 sets of study models
Rejection criteria included: was deemed to be adequate and sufficiently representa-
tive in relation to the variance of the larger group and to
N gross restorations, build-ups, crowns, onlays, Class II
the size of error regarded to be of clinical significance,
amalgam or composite restoration that affect the
which was judged to be half the size of TSD (1.5 mm)
tooth’s mesiodistal diameter;
which Proffit6 considered of potential clinical signifi-
N congenitally missing teeth and impacted teeth.
cance. All the teeth were measured twice for the two
The mesiodistal diameter tooth sizes were measured methods, with a week between the measurements. An
from first molar to first molar at the level of the contact inter-examiner calibration involved three additional
 236 Othman and Harradine Scientific Section JO December 2007

examiners – an experienced orthodontist (NWTH) and significant difference found between the two sets of
two senior trainees (CD and SD) with four years of measurements, however the P value for time was close
previous orthodontic training. These three examiners to statistical significance (P50.058) and again showed a
also measured the 20 study models twice using the small reduction in time with the second set of measure-
HATS method only, to determine intra- and inter- ments. Table 1c contains the analysis of systematic
examiner systematic error and to compare random differences between four examiners using the HATS
errors for that method. method. Statistically significant differences were found
between the four examiners for the three total arch
Statistical analysis measurements, but not for the anterior arch measure-
The distribution of data was evaluated for normality. ments or the time. Results also reflected the results for
For assessment of the systematic error, repeated SAO in Table 1b, in that there was no within-operator
measures analysis of variance (ANOVA) and the systematic bias for the HATS method for any of the
Greenhouse–Geisser approximation were used to test operators. The mean figures in Table 1c are the averages
statistical significance. Greenhouse–Geisser is a stan- of the duplicated measurements.
dard method of dealing with sphericity with the
Random error
assumption that each of the examiners was related to
each other in the same way. The paired-sample t-test Random errors are given in Table 2 for Odontorule and
was used to evaluate the systematic error and the Table 3 for HATS for observer SAO. The error variance
differences in timings for the two methods of measure- is a high percentage of the total variance for all measures
ment. Random error was calculated in terms of the and for both methods of measurement. The total variance
standard deviation of the differences in replicate for the sample of 150 was, by chance, much higher than
measurements as advocated by Houston.15 The variance the total variance for the sample of 20 used for the
of the difference between two replicate measurements is duplicate measurements. A sample size of 20 was judged
double that of a single measurement, so the variance of to be completely sufficient for all analytical purposes
the differences must be halved to give a correct estimate except for the comparison of error variance to total
of the error for a single measurement. This measure was variance. The total variance of the randomly chosen 20
preferred to the root mean square error (as advocated by subjects was unpredictably larger by chance than for the
Dahlberg,16 and which is still frequently employed), sample of 150, so the variance of the full 150 sample was
because it avoids the possibility of any systematic bias the better choice for the comparison of error variance to
affecting the assessment of random error. Dahlberg’s total variance. Extending the duplication of measure-
formula is only accurate if there is no systematic bias. ments to a number greater than 20 would only stand a
The analysis in the present study included the percen- small chance of increasing the reliability of all other values
tage of the total sample variance that consists of error in the reproducibility analysis. Table 4(a–d) therefore
variance (the variance of replicate measurements), contains the random error analyses for all four observers
because Houston makes the crucial point that the for the HATS method, using this more representative
potential effect of random error on interpretation of complete sample variance for comparison with the error
results can only be properly estimated in relation to the variance. The percentages of error variance were corre-
variance from all sources in a representative sample. spondingly lower than in Table 3, but still high.

Results Systematic error (bias) between the two methods

By inspection, all the data were demonstrated to come Table 5 shows that the mean differences for tooth-size
from a normally distributed population and parametric discrepancy are very small and not statistically signifi-
tests were therefore used. cant, but the speed of the measurement and analysis for
the two methods was very different, the HATS method
Systematic error being much faster.

The systematic error (bias) of the principal investigator

(SAO) for the Odontorule is detailed in Table 1a. There
Discussion
was a small but statistically significant difference Systematic error
between the two means for the upper total correction,
lower total correction and the time. Table 1b shows the Table 1(a,b) lists the systematic error results of the
same data for the HATS method where there was no two methods of measurement performed on the same
 JO December 2007 Scientific Section Tooth-size discrepancy and Bolton’s ratios 237

casts twice by the same investigator. The HATS results to the study. This familiarization factor is of potential
showed no significant differences between the means significance for the occasional user in a clinical setting.
although the mean reduction in measurement time of The Odontorule results also show statistically signifi-
0.21 minutes or 12 seconds approached significance. cant differences in the mean total correction values. The
For the Odontorule, the mean reduction in measure- mean differences are approximately 0.6 mm and are
ment time of 42 seconds for the second measurement therefore small in terms of clinical significance – Proffit6
was statistically significant. Both these results suggest felt that a discrepancy of ,1.5 mm is rarely of
that a process of familiarization was still occurring significance. The same values when measured with
during this part of the study, in spite of fairly extensive HATS show a similar trend but to a smaller and non-
use of both methods by the principal investigator prior significant extent. Table 1c shows statistically significant

Table 1 (a) Intra-examiner reproducibility (systematic error): Odontorule. Observer SAO (n520).

95% Confidence intervals

Descriptive Mean time 1 Mean time 2 Mean difference Lower Upper P value

Total Bolton ratio (%) 91.23 91.90 –0.68 –1.13 –0.22 0.054
Upper total correction (mm) –0.29 0.45 0.72 –1.26 –0.18 0.007*
Lower total correction (mm) 0.07 –0.49 0.56 0.04 1.07 0.012*
Anterior Bolton ratio (%) 78.07 78.80 –0.73 –1.40 –0.17 0.088
Upper anterior correction (mm) 0.80 0.94 –0.14 –0.53 0.25 0.216
Lower anterior correction (mm) –0.60 –0.77 0.17 –0.23 0.54 0.270
Time (minutes) 9.53 8.82 0.70 0.34 1.06 0.001*

*P,0.05.

Table 1 (b) Intra-examiner reproducibility (systematic error): HATS. Observer SAO (n520).

95% Confidence intervals

Descriptive Mean time 1 Mean time 2 Mean difference Lower Upper P value

Total Bolton ratio (%) 91.18 91.58 20.40 21.15 0.03 0.135
Upper total correction (mm) 20.40 0.29 20.43 21.00 0.14 0.134
Lower total correction (mm) 0.13 20.26 0.40 20.13 0.91 0.135
Anterior Bolton ratio (%) 78.15 78.30 20.15 20.79 0.50 0.639
Upper anterior correction (mm) 0.54 0.66 20.12 20.50 0.30 0.389
Lower anterior correction (mm) 20.41 20.51 0.10 20.21 0.41 0.393
Time (minutes) 3.97 3.46 0.21 20.00 0.43 0.058

*P,0.05.

Table 1 (c) Intra-examiner reproducibility (systematic error): HATS (n520).

P value for difference

Descriptive Mean (SAO) Mean (NWTH) Mean (CD) Mean (SD) between people

Total Bolton ratio (%) 91.38 91.21 91.15 91.71 0.014*

Upper total correction (mm) 0.07 20.10 20.16 0.43 0.014*
Lower total correction (mm) 20.07 0.09 0.15 20.40 0.011*
Anterior Bolton ratio (%) 78.2 78.04 78.01 77.99 0.760
Upper anterior correction (mm) 0.60 0.42 0.48 0.46 0.783
Lower anterior correction (mm) 20.46 20.34 20.37 20.36 0.735
Time (minutes) 3.56 3.75 4.73 2.99 0.077

*P,0.05.
 238 Othman and Harradine Scientific Section JO December 2007

Table 2 Random error: Odontorule. Observer SAO (n520).

Total correction Anterior correction

Total ratio Upper Lower Anterior ratio Upper Lower Time

Total variance 1.82 2.06 1.64 2.21 0.87 0.68 0.68

Mean difference 20.68 20.72 0.56 20.73 20.14 0.17 0.70
SD of differences 0.98 1.16 1.11 0.92 0.84 0.83 0.77
Variance/2 0.48 0.67 0.61 0.42 0.36 0.34 0.29
Error variance (%) 26.23 32.42 37.54 9.28 40.68 0.32 43.57
Correlation coefficients (r) 0.746 0.665 0.619 0.658 0.605 0.545 0.508

systematic differences in the mean measurements method of measuring is probably inappropriate. The
obtained by the four operators. These mean differences results from the full sample of 150 revealed a substan-
were very small (less than 0.5% and less than 0.5 mm) tially larger total variance and the expert statistical
and were again confined to the total arch measurements. advice was that the larger sample is a more valid
Nevertheless, the existence of any systematic error indicator of total variance in the orthodontic popula-
suggests that considerable familiarity with these techni- tion. The results using this total variance are in Table 4
ques is required before there is stability of point and show that for the main examiner (SAO, Table 4a),
identification and that occasional use of this analysis is this estimate of the percentage random error was much
not appropriate. Inter-operator errors were not analysed closer to the 10% recommended by Midtgård et al.17 but
for the manual Odontorule method because it had remains higher than is desirable for a robust measure-
already become apparent that the substantial additional ment method.
time required for this method, with no evidence of The inter-examiner reproducibility was then assessed
greater reproducibility, made this a method which could to see whether this percentage of error variance was
not be recommended for clinical use. particular to the main examiner. It can be seen in
Table 4(b–d) that for all examiners the error variance
Random error was a higher percentage of the total variance than
advocated by Midtgård et al.17 and by Houston.15 There
The results for the main examiner (SHO) for the two were some differences between the examiners, but none
methods are shown in Tables 2 and 3, and they are of these differences in random error was statistically
similar for both methods. The standard deviation of significant as assessed by Greenhouse–Geisser. Two
replicate measurements is of the order of 1 mm for investigators (CD and SD) were much less familiar with
correction and 1% for ratios. These standard deviations the HATS method, but their random error was similar
are significant, being more than half the size of TSD to the main investigator (SAO) who was significantly
(1.5 mm) which Proffit6 considered of significance. It is more familiar with this process. NWTH was familiar
also important to place the error variance in the context with the method and over a longer period of time and
of the total variance of the sample. Midtgård et al.17 had lower random error values, which were within the
suggested that the error variance should not exceed 3% recommended 10%, but higher than the ideal 3%. The
of the total variance, and if it exceeded 10% the applied results suggest that experience may reduce random error

Table 3 Random error: Hats. Observer SAO (n520).

Total correction Anterior correction

Total ratio Upper Lower Anterior ratio Upper Lower Time

Total variance 1.61 1.71 1.42 2.50 0.90 0.53 0.15

Mean difference 20.40 20.43 0.40 20.15 20.12 0.09 0.21
SD of differences 1.16 1.22 1.12 1.37 0.86 0.66 0.47
Variance/2 0.67 0.75 0.63 0.94 0.37 0.22 0.11
Error variance (%) 41.58 43.85 43.96 37.88 40.93 41.25 74.85
Correlation coefficients (r) 0.505 0.561 0.560 0.633 0.609 0.609 0.254
 JO December 2007 Scientific Section Tooth-size discrepancy and Bolton’s ratios 239

Table 4 (a) Random error: Hats. Observer SAO. Using complete sample variance (n5150) for error variance %.

Total correction Anterior correction

Total ratio Upper Lower Anterior ratio Upper Lower

Complete sample variance 3.42 3.55 2.96 5.25 1.80 1.07

Mean difference 20.40 20.43 0.40 20.15 20.12 0.09
SD of differences 1.16 1.22 1.12 1.37 0.86 0.66
Variance/2 0.67 0.75 0.63 0.94 0.37 0.22
Error variance (%) 19.5 21.1 21.28 17.90 20.55 20.56

Table 4 (b) Random error: Hats. Observer NWTH. Using complete sample variance (n5150) for error variance %.

Total correction Anterior correction

Total ratio Upper Lower Anterior ratio Upper Lower

Complete sample variance 3.42 3.55 2.96 5.25 1.80 1.07

Mean difference 0.22 0.23 20.21 0.17 0.11 20.02
SD of differences 0.71 0.72 0.65 0.85 0.50 0.43
Variance/2 0.26 0.26 0.21 0.36 0.13 0.09
Error variance (%) 7.60 7.32 7.09 6.85 7.22 8.40
Correlation coefficients (r) 0.838 0.852 0.853 0.851 0.852 0.814

Table 4 (c) Random error: Hats. Observer CD. Using complete sample variance (n5150) for error variance %.

Total correction Anterior correction

Total ratio Upper Lower Anterior ratio Upper Lower

Complete sample variance 3.42 3.55 2.96 5.25 1.80 1.07

Mean difference 20.25 20.25 0.23 20.17 20.09 0.07
SD of differences 0.73 0.77 0.72 1.15 0.72 0.56
Variance/2 0.26 0.31 0.26 0.66 0.26 0.16
Error variance (%) 7.60 9.73 9.78 12.57 14.44 14.95
Correlation coefficients (r) 0.786 0.771 0.772 0.737 0.722 0.723

Table 4 (d) Random error: Hats. Observer SD. Using complete sample variance (n5150) for error variance %.

Total correction Anterior correction

Total ratio Upper Lower Anterior ratio Upper Lower

Complete sample variance 3.42 3.55 2.96 5.25 1.80 1.07

Mean difference 0.14 0.14 20.14 0.17 0.10 20.08
SD of differences 0.89 0.91 0.83 1.39 0.83 0.64
Variance/2 0.40 0.42 0.35 0.98 0.34 0.20
Error variance (%) 11.69 11.83 11.82 18.67 18.87 18.69
Correlation coefficients (r) 0.823 0.826 0.828 0.683 0.689 0.689
 240 Othman and Harradine Scientific Section JO December 2007

to a worthwhile extent, but not to a level where between the HATS and the Vernier callipers was
confidence can be placed in a single measurement. r50.825. However, there was no separate test for
Statistical analysis of this suggestion is complicated by random error with either method in their study.
the difficulty in quantifying with validity the relevant Shellhart et al.8 studied the reliability of Bolton’s
experience in a group of four operators. tooth-size analysis when applied to crowded dentitions
These results are clearly important in relation to the using needle-pointed dividers and the Boley gauge. For
assessment of a single patient. Great caution should be 14 of the 16 measures, there was no statistically
exercised before instituting an intervention on the basis significant difference. Random error was estimated by
of one measurement of the Bolton discrepancy. correlation coefficients. These varied very greatly from a
Confidence in the calculation of discrepancy is particu- reasonable correlation of r50.79 to a very low figure of
larly important if the resulting intervention is reduction r520.15 for intra-investigator errors. Intraclass coeffi-
of tooth width by interdental stripping or extraction. To cients for measurements made by four investigators
reduce the random error for both methods of measure- ranged from 0.80 to 0.29. Many of these values are
ment explored in this study, a clinician is strongly therefore very much lower than would be considered
advised to measure the same study models three or four desirable for a good method of measurement. The
times and then average the values obtained before authors agreed with this view and stated that, ‘If a
committing to any active intervention. clinician’s repeatability of the Bolton analysis is average,
calculations of tooth-size discrepancy should be viewed
Error in relation to tooth irregularity as ¡2.2 mm.’ This recommended confidence level is
very large in relation to a clinically significant TSD and
Locating contact points on a crowded dentition is
their conclusion begs the question as to what method
difficult. The sample in this present study consisted of a
should, in their view, actually be used to decide on
variety of malocclusions with a range of crowding.
therapeutic intervention.
Shellhart et al.8 found that every investigator made at
Zilberman et al.9 also reported that measurement with
least one error in measurement that was greater than a
clinically significant value for the tooth-size excess when digital callipers on plaster models showed better
measuring Bolton discrepancies on crowded dentitions reproducibility than measurements on virtual computer-
(at least 3 mm of crowding) with a Boley gauge and ized models (OrthoCAD). Importantly, the repeated
needle-point dividers. It would be possible to take study measures of the total tooth-size widths were evaluated,
casts in the middle of treatment for analysis once but not Bolton’s ratios or the discrepancies. They found
alignment had been achieved and in very crowded both random and systematic errors were very small and
dentitions, this is advisable. clinically insignificant. The error of the sum of the tooth
widths is, however, likely to be much smaller than the
error in the calculated Bolton’s ratios or correction in
Comparison of errors between methods of measurement
millimetres, because the sum of tooth widths is a much
The present study indicated that there were no larger absolute figure. Direct comparison of the
differences between the two methods of measurement. reproducibility of Zilberman et al.9 with the present
This result is in agreement with that reported by study is therefore not reliable. As has been mentioned,
Tomassetti et al.7 who compared the HATS system some well-known studies did not report the measure-
and Vernier callipers. Their correlation coefficient ment error at all11 or very inadequately.12–14

Table 5 Comparison of mean results for the Odontorule and HATS methods (n520). Observer SAO. Paired t-test.

95% Confidence intervals

Descriptive Mean Odontorule Mean HATS Mean difference Lower Upper P value

Total Bolton ratio (%) 91.23 91.18 20.05 20.56 0.45 0.825
Upper total correction (mm) 20.29 20.40 0.14 20.41 0.68 0.603
Lower total correction (mm) 0.07 0.13 0.05 20.46 0.56 0.825
Anterior Bolton ratio (%) 78.07 78.15 0.07 20.59 0.84 0.774
Upper anterior correction (mm) 0.80 0.54 20.26 20.64 0.12 0.168
Lower anterior correction (mm) 20.60 20.41 0.18 20.17 0.61 0.341
Time (minutes) 8.93 3.49 25.45 25.88 25.01 ,0.001
 JO December 2007 Scientific Section Tooth-size discrepancy and Bolton’s ratios 241

Differences in timing reducing error. Analysis on an occasional basis is

not advisable.
The present study noted that the HATS method was
quicker than the Odontorule. This average difference of
5.45 minutes is significant both statistically and clini-
cally. The comparative times for the four examiners
Contributors
(Table 1c) show a range of mean times for the HATS Siti Adibah Othman was responsible for material and
method which was much smaller (1.75 minutes) than the data collection, conducting the experiment, data analy-
difference between the two methods. These results are sis and writing of the article. Mr Nigel Harradine was
similar to those of Tomassetti et al.7 This part of the responsible for study design, drafting of the article,
study strongly supported the abandonment of manual contributed to the writing of the article, data analysis
methods of measurement and calculation in favour of and interpretation, critical revision and final approval
direct electronic entry of measurements into a software of the published version. Mr Nigel Harradine is the
package. The two methods gave essentially the same guarantor.
mean values (Table 5) but the HATS method was much
faster, potentially less prone to blunder or fatigue
problems and was of comparable or better reproduci- Acknowledgements
bility. Both methods are technically easy to use. The
The authors would like to express their thanks to
Odontorule costs approximately £30 and the HATS
Christian Day and Scott Deacon for their involvement
software approximately £250. Electronic callipers are
in the assessment of inter-examiner reproducibility and
used with both methods and cost approximately £80.
to Mrs Rosemary Greenwood for her thorough and
expert guidance throughout the statistical analysis.
Strengths and weaknesses of this study
This study benefits from a sample which is very
representative of clinical orthodontic practice. The
References
statistical analysis importantly measures both random 1. Ballard ML. Asymmetry in tooth size: a factor in the
error and systematic error and relates that error directly etiology, diagnosis and treatment of malocclusion. Angle
to the dimensional changes of clinical significance. This Orthod 1944; 14: 67–71.
enables important clinical conclusions, namely that 2. Neff CW. Tailored occlusion with the anterior coefficient.
random error is high, so repeat measures are needed on Am J Orthod 1949; 35: 309–14.
an individual case and that infrequent use of this method 3. Steadman SR. The relation of upper anterior teeth to lower
is not appropriate. The study shows that the manual anterior teeth as present on plaster models of a group of
method of measurement no longer has a clinical role, acceptable occlusions. Angle Orthod 1952; 22: 91–97.
being much slower and with the same reproducibility. 4. Lundstrom A. Intermaxillary tooth width ratio and tooth
Such a study could be improved by increasing the alignment and occlusion. Acta Odontol Scand 1954; 12: 265–
number of operators and by measuring on more than two 92.
occasions. This could provide data on how much 5. Bolton WA. Disharmony in tooth size and its relation to the
experience with this method is required to eliminate any analysis and treatment of malocclusion. Angle Orthod 1958;
tendency to drift in point identification and therefore 28: 113–30.
statistical bias. It might also confirm whether, and by how 6. Proffit WR. Contemporary Orthodontics, 3rd edn. St Louis:
much, greater experience also reduces random errors. Mosby, 2000, 170.
7. Tomassetti JJ, Taloumis LJ, Denny JM, Fischer JR Jr. A
comparison of 3 computerized Bolton tooth-size analyses
Conclusions with a commonly used method. Angle Orthod 2001; 71: 351–
57.
N The HATS and Odontorule methods had similar 8. Shellhart WC, Lange DW, Kluemper GT, Hicks EP Kaplan
reproducibility, but the HATS method was signifi- AL. Reliability of the Bolton tooth size analysis when
cantly quicker. applied to crowded dentitions. Angle Orthod 1995; 65: 327–
N Single estimations of TSD should be treated with 34.
great caution and replicate measurements are advised 9. Zilberman O, Huggare JA, Parikakis KA. Evaluation of the
if active clinical intervention is planned. validity of tooth size and arch width measurements using
N There is evidence to suggest that significant experience conventional and three-dimensional virtual orthodontic
with measuring TSD has a worthwhile effect in models. Angle Orthod 2003; 73: 301–6.
 242 Othman and Harradine Scientific Section JO December 2007

10. Othman SA, Harradine NW. Tooth-size discrepancy and 14. Bernabé E, Major PW, Flores-Mir C. Tooth-width ratio
Bolton’s ratios: a literature review. J Orthod 2006; 33: 45–51. discrepancies in a sample of Peruvian adolescents. Am J
11. Freeman JE, Maskeroni AJ, Lorton L. Frequency of Orthod Dentofacial Orthop 2004; 125: 361–65.
Bolton tooth-size discrepancies among orthodontic 15. Houston WJ. The analysis of errors in orthodontic
patients. Am J Orthod Dentofacial Orthop 1996; 110: 24–27. measurements. Am J Orthod 1983; 83: 382–90.
12. Crosby DR, Alexander CG. The occurrence of tooth size 16. Dahlberg G. Statistical Methods for Medical and Biological
discrepancies among different malocclusion groups. Am J Students. New York: Interscience publications, 1940.
Orthod Dentofacial Orthop 1989; 95: 457–61. 17. Midtgård J, Björk G, Linder-Aronson S. Reproducibility of
13. Araujo E, Souki M. Bolton anterior tooth size discrepancies cephalometric landmarks and errors of measurements of
among different malocclusion groups. Angle Orthod. 2003; cephalometric cranial distances. Angle Orthod 1974; 44: 56–
73: 307–13. 61.