RESEARCH AND EDUCATION

Accuracy in the digital workflow: From data acquisition

to the digitally milled cast
George K. Koch,a German O. Gallucci, DMD, Dr Med Dent, PhD,b and Sang J. Lee, DMD, MMScc

Computer-aided design and ABSTRACT

computer-aided manufacturing Statement of problem. The accuracy of digital impressions and the milling of implant crowns
(CAD/CAM) has revolution- greatly influence the clinical outcome of implant restorations.
ized industry and the manu-
Purpose. The purpose of this in vitro study was to calculate the propagation of error in the process
facture of goods. CAD/CAM
of milling an implant crown.
was first incorporated into
dentistry in the 1980s and is Material and methods. Thirty digitally milled casts made directly from a reference model were
now used for digital impres- prepared. The casts were scanned with a laboratory scanner, and 30 standard tesselation language
(STL) datasets from each group were imported to inspection software. In each analysis, STL datasets
sions and the in-office milling
were aligned by a repeated best fit algorithm, and 18 specified contact locations of interest were
of restorations.1-3 A digital measured in mean volumetric deviations. The master reference dataset was aligned 30 times to the
impression allows clinicians to master reference dataset to determine the software variation. The reference datasets were aligned
visualize the preparation and to the master reference dataset to determine the scanner variation. The milled cast datasets were
design the prostheses so that aligned to the master reference dataset to determine the milling variation. The 18 specified contact
they can evaluate the prepara- locations of interest were pooled by cusps, occlusal ridge/fossae, interproximal contacts, facial/
tions before the laboratory lingual aspect, and implant position. The pooled areas were statistically analyzed by comparing
4,5 each group with the reference model to investigate the mean volumetric deviations accounting for
processes.
accuracy and standard deviations for precision.
Digital impressions offer
many advantages over con- Results. Software and scanner variation were negligible. Variations in the milled models resulting
from software and scanner error exhibited statistical significance (P<.001). Software, scanner, and
ventional impressions. They
milling error were shown to propagate through the digital workflow to the milled model.
provide virtual diagnostic
assessment of restorative pla- Conclusions. The pooled locations may describe the reliability of the milling process as it applies to
nning, reduced distortion of specific anatomic locations on the tooth. (J Prosthet Dent 2016;115:749-754)
impression materials, patient
acceptance, and potential cost and time effectiveness.6 In a digital workflow for the accurate fabrication of
Despite these advancements, at present, digital im- dental restorations, discrepancies in precision and
pressions cannot fully replace the conventional impres- trueness have been noted.9 Precision describes how close
sion for restorative procedures because of limited repeated measurements are to each other.10 Trueness
technologic resources and practicability in the restorative refers to the deviation between the measured di-
workflow.7,8 Determining how digital technology can be mensions and actual dimensions of the object. Dis-
applied in a digital workflow is essential in the course of crepancies can be incorporated in each step of a digital
treatments. workflow.

Supported in part by the ITI Foundation, Basel, Switzerland (research grant 733-2010).
a
Graduate student, Harvard School of Dental Medicine, Boston, Mass.
b
Associate Professor and Chair, Department of Restorative Dentistry and Biomaterials Science, Harvard School of Dental Medicine, Boston, Mass.
c
Instructor, Department of Restorative Dentistry and Biomaterials Science, Harvard School of Dental Medicine, Boston, Mass.

THE JOURNAL OF PROSTHETIC DENTISTRY 749

750 Volume 115 Issue 6

Clinical Implications
The propagation of error in a digital workflow
influences the accuracy in fabricating a physical
cast. Direct digitization by means of the intraoral
scanner had less systematic error than the physical
cast fabrication step by means of the milling
process from an intraoral scan.

The sources of inaccuracy can come from the direct

3-dimensional (3D) data acquisition by digital scan-
ners.11-13 A transformation error can occur when a direct
digital data acquisition is transferred to CAD software.14-16
In the process of milling physical models or restorations,
inaccuracy can arise from milling and 3D printing
because of the limited capability to generate the details of
dental restorations and casts. Inaccuracy throughout the
processes can be formulated by error propagation and
calculation.17
The purpose of this in vitro study was to investigate
the propagation of error in the process of a digital
workflow in the fabrication of dental restorations. By
recognizing the sources of error and their propagation
throughout the system, the process can be improved to
generate better laboratory and clinical outcomes. The null
hypothesis was that no significant difference in accuracy
would be found in the digital workflow from a digital
acquisition to a milled cast.
MATERIAL AND METHODS
A scannable customized maxillary model (Models Plus)
retaining a single implant (Bone Level, Regular CrossFit;
Straumann) located in the maxillary left second premolar
position was used as a master model. A scannable abut-
ment (Scanbody; Straumann AG) was attached, and the
master model was digitally scanned 30 times with an
intraoral scanner (Cadent i-Tero; Align Technology Inc).
The standard tesselation language (STL) datasets were
then transferred to a custom milling center (Carstadt, New
Jersey), where 30 digitally milled casts were fabricated.
The master model and 30 digitally milled casts were
scanned with a reference scanner (Lava Scan ST; 3M
ESPE) with scannable abutments on the implants to
digitize the position and angulation of the implants. The
surface scan included the area from the mesial of the
maxillary left premolar to the distal of the maxillary left
first molar. Thirty STL datasets from models digitally
milled with a reference scanner were paired to 30 STL
datasets from the intraoral scanner. All the paired STL
datasets from a master model, the virtual casts from the
intraoral scanner, and the digitally milled casts were
imported into inspection software (Geomagic Qualify
12.0; Geomagic).
Figure 1. Steps in the digital workflow. (A) Master versus master, (B) master
versus intraoral, (C) intraoral versus milled, and (D) master versus milled.
All alignments were performed using a repeated best
fit algorithm based on the selected surfaces of the
maxillary left first premolar and molar. The scans were
first oriented in a buccal and lingual view. A polygonal
selection tool was used to select STL datasets on the
buccal and lingual surfaces of the maxillary left first
premolar and molar. Next, a repeated best fit algorithm
aligned the scans into the same 3D coordinate system.
Once aligned into the same 3D coordinate system as the
master scan, the 18 specified contact locations of interest
were calculated by the inspection software in mean
volumetric deviations. The mean volumetric deviations
can have positive or negative values in relation to the
reference dataset. By aligning all scans into the same 3D
coordinate system as the master scan, the inspection
software was able to measure the same 18 contact loca-
tions in all scans.
Figure 1 represents the analysis of the different groups
to investigate the errors in each step of the digital
workflow. The master reference dataset was aligned to
the master reference dataset 30 times to determine the
software variation (master versus master). The 30 refer-
ence datasets were aligned to the master reference
dataset to determine the variation caused by 3D data
acquisition by an intraoral scanner (master versus
intraoral). The variation in the comparison of the 3D data
acquisition and the data for milled casts provided the

THE JOURNAL OF PROSTHETIC DENTISTRY Koch et al

June 2016
Figure 2. Three-dimensional differences between test and reference STL
(standard tesselation language) file. Generated following best fit align-
ment by 3D inspection software. Color-coded scale represents discrep-
ancy of matching (mm).
errors throughout the fabrication process (intraoral
versus milled). Finally, the 30 milled cast datasets were
aligned to the master reference dataset to determine the
milling variation which represents the overall accuracy
throughout a digital workflow (master versus milled).
The divergences in the x-, y-, and z- axes between
each reference and test dataset at 18 specified contact
locations were measured. The 18 specified contact lo-
cations of interest were pooled by cusps, occlusal ridge/
fossae, interproximal contacts, facial/lingual aspect, and
implant position. The pooled areas were statistically
analyzed by comparing each group with the reference
model to investigate the mean volumetric deviations.
The average mean volumetric deviation between refer-
ence and test datasets in a group represents the accu-
racy and standard errors for precision. The 18 specified
areas of interest were located in the horizontal and
vertical axes as follows: the distal and mesial inter-
proximal contact points of the implant site, top of the
scannable implant abutment, buccal and mesiobuccal
cusps, palatal and mesiolingual cusps, and distal and
mesial fossae of the first premolar and first molar
(Fig. 2). The size of specified locations of analysis was
0.5×0.5 mm, and the divergences were measured in a
uniform position on the reference coordinate axes. For
further evaluations, the deviations between the refer-
ence dataset and test datasets of each single measure-
ment point were exported to software (SAS 9.3; SAS
Institute Inc).
A statistical analysis was performed with software
(SAS 9.3; SAS Institute Inc) to investigate the variations
caused at each step of the digital workflow. A linear
mixed model was applied to analyze the data for a
repeated measurement of each reference point. In the
model, each sampled point has its own means in each
measurement, including a random intercept. Outliers
were excluded from the analysis based on residual
analysis. Group comparisons were obtained through an
indicator variable. Multiple comparison adjustment was
also applied to avoid type I error inflation.
Koch et al
751
Table 1. Mean volumetric deviations (mm) by test group (outliers
excluded)
Group Mean Minimum Median Maximum P
Master versus master 0.000 ±0.001 -0.001 0.001 0.040 .465
Master versus intraoral -0.001 ±0.021 -0.095 -0.005 0.086 .669
Intraoral versus milled -0.008 ±0.098 -0.095 -0.010 1.631 <.001
Master versus milled -0.010 ±0.100 -0.120 -0.011 1.724 <.001
RESULTS
After the repeated best fit alignment, the mean volu-
metric deviations were calculated by test group (Table 1)
and pooled location (Table 2). Accuracy is presented as
the mean volumetric deviation. Precision is presented as
the standard deviation (SD). A positive or negative mean
volumetric deviation correlates with the test dataset be-
ing over or under the reference dataset. Areas of over-
reduction in the milled models are therefore presented as
negative volumetric deviations.
On the basis of the mean volumetric deviations and
the SD of all 18 datasets from each group, master versus
master showed the fewest deviations, followed by master
versus intraoral, intraoral versus milled, and master
versus milled. In terms of maximal positive and negative
deviations, master versus milled showed the highest di-
vergences, followed by intraoral versus milled, master
versus intraoral, and master versus master. Software and
scanner variation (master versus master and master
versus intraoral) were negligible. Milled models from
digital impressions had comparable accuracy digital
models. The test groups master versus milled and
intraoral versus milled exhibited statistical significance
(P<.001). Table 1 excludes the 5 implant points as outliers
in all 4 test groups.
In Table 2, on the basis of the SD of all points from
each group, the scanner (master versus intraoral)
showed the greatest precision along the facial/lingual
aspect. This was followed by the occlusal ridge and
fossa, interproximals, cusps, and implant. In terms of
mean volumetric deviations, positive deviations in the
master versus intraoral group correlated with negative
deviations in the intraoral versus milled group and vice
versa. This had an additive/subtractive effect, seen in the
master versus milled group, wherein the absolute value
of the mean volumetric deviation was less than the
corresponding value in either the master versus
intraoral, intraoral versus milled, or both. This same
effect was seen in the SD of the master versus milled
group. The SD of the interproximals, cusps, facial/
lingual aspect, and occlusal ridge and fossa was less
than either the master versus intraoral, intraoral versus
milled, or both. There was one exception; the implant
group had a master versus milled SD (0.170 mm) that
exceeded its corresponding value in intraoral versus
milled (0.162 mm).
THE JOURNAL OF PROSTHETIC DENTISTRY
752 Volume 115 Issue 6

Table 2. Mean volumetric deviations (mm). Pooled by anatomic location (outliers included)
No. of Points Master Master versus Intraoral versus Master versus
Group in Group versus Master Intraoral Milled Milled
Interproximals 2 0.001 ±0.001 0.006 ±0.023 -0.005 ±0.017 0.000 ±0.013
Cusps 4 0.005 ±0.005 0.008 ±0.025 -0.044 ±0.029 -0.038 ±0.023
Implant 5 0.001 ±0.004 -0.006 ±0.040 0.019 ±0.162 0.014 ±0.170
Facial/lingual aspect 4 0.001 ±0.001 -0.012 ±0.008 0.026 ±0.015 0.014 ±0.013
Occlusal ridge and fossa 3 0.004 ±0.005 0.002 ±0.017 -0.033 ±0.029 -0.032 ±0.027

DISCUSSION produce a systematic error. Flugge et al8 found that

digitizing the molar areas with the iTero was less precise,
The purpose of this study was to calculate the propaga-
possibly because of the complex angled surfaces of the
tion of error in the process of milling an implant crown.
molars and the undercuts on surrounding teeth.
This was accomplished by performing a repeated best fit
In the master versus intraoral group a Lava ST scan
alignment and calculating the mean volumetric deviation
was used as the reference and i-Tero was used as the test.
between reference and test datasets at 18 points of in-
In the intraoral versus milled group, a i-Tero scan was
terest. The anatomic locations of the points were chosen
used as the reference and Lava ST was used as the test.
based upon their clinical applications. Measurements
The master versus milled group, however, used both a
were made at 18 predetermined locations, which allowed
Lava ST scan for both the reference and the test. The
for standardization and reproducibility. However, using
master versus milled group can be regarded as the most
preset fixed measurement points did not account for
reliable in determining the accuracy of the milling process
discrepancies in all other nonmeasured locations.
at specific anatomic locations. In terms of mean volu-
Repeated best fit alignments have been previously
metric deviations, positive deviations in the master
used in several other studies as an approach to 3D
versus intraoral group correlated with negative deviations
comparison.4,5,14,15 Rudolph et al11 evaluated 3 different
in the intraoral versus milled group and vice versa. This
dental CAD-systems. Large deviations occurred in areas
could be due to using an i-Tero scan as the reference
with strong changes of curvature. This can be related to
versus the test. An alternative explanation would be that
the density of the point-clouds in those areas.12,13 For the
the milling process overreduced in areas where the
purposes of this study, a repeated best fit alignment was
scanner had positive deviations and vice versa.
performed using preselected surfaces on the facial and
The former explanation is much more likely, because
lingual aspects of the molar and premolar. These areas
error can be shown to propagate through the digital
have the least amount of curvature which should result in
workflow. Every measurement includes a component of
the least amount of deviation. However, the possibility
error. Typically error is represented by the SD. In the
exists that by limiting the alignment to these surfaces, the
process of a digital workflow, error comes from many
resulting mean volumetric deviations on the facial/lingual
sources including the software, scanner, and milling
aspect could be underreported. If this was the case, one
process. Variations in individual scans result in differing
would expect the facial/lingual aspect to have the
milled casts. Master versus master contains software er-
smallest values for mean volumetric deviation (Table 2).
ror. Master versus intraoral is affected by both software
This was not seen in master versus intraoral, intraoral
and scanner variation. Intraoral versus milled contains
versus milled, or master versus milled. The SD for the
software and milling error. Lastly, master versus milled is
facial/lingual aspect, however, was the smallest and may
affected by software, scanner, and milling variation. This
be attributable at least in part to the alignment method.
is supported by Table 1, wherein master versus master
An alternative explanation is that the facial and lingual
showed the least deviations (0.000 ±0.001 mm), followed
aspect of the molar and premolar have the least complex
by master versus intraoral (-0.001 ±0.021 mm), intraoral
morphology and the milling process is very reliable in
versus milled (-0.008 ±0.098 mm), and master versus
these areas. By limiting the alignment to these selected
milled (-0.010 ±0.100 mm).
areas, extraneous datasets were not used in the align-
One of the common formulas used by engineers is the
ment process. Extraneous datasets were available, how-
variance formula, sf=√(sx2+sy2+sz2).17
ever, following the alignment for calculation of the mean
The formula provides a simple approach to calculating
volumetric deviation.
the propagation of uncertainty. The formula is limited in
One potential source of systematic error is that the
its application because random errors in measurement
study used 2 different scanning technologies. The i-Tero
are assumed to be independent of one another. When
system uses a process called stitching wherein single
applied to the results from Table 1, variation can be
images are assembled for a virtual model.8 The stitching
shown to propagate through the digital workflow.
algorithm of the i-Tero system is unknown and may

THE JOURNAL OF PROSTHETIC DENTISTRY Koch et al

June 2016
sx =√ðssoftware 2 +sscanner 2 +smilling 2 Þ
(Table 1): =√(0.0012+(0.0212-0.0012)+(0.0982-0.0012))
gives an anticipated SD in the master versus milled group
of 0.100 mm; the actual SD was 0.100 mm. Software,
scanning, and the milling process all contributed error
(0.001, 0.021, and 0.098 mm, respectively) to the milled
product (0.100 mm).
Given that variation propagates throughout the digi-
tal workflow, one would expect to find a master versus
milled SD (Table 2) that exceeds both master versus
intraoral and intraoral versus milled in all anatomic lo-
cations. In Table 2, the master versus milled SD of the
interproximals, cusps, facial/lingual aspect, and occlusal
ridge and fossa were less than the master versus
intraoral, intraoral versus milled, or both. Only the
implant group had a master versus milled SD (Table 2)
that exceeded both master versus intraoral and intraoral
versus milled. This suggests that the variations in master
versus intraoral and intraoral versus milled were insig-
nificant at the interproximals, cusps, facial/lingual aspect,
and occlusal ridge and fossa. These small variations may
simply be due to the use of 2 different scanning tech-
nologies in these areas.
The implant group (Table 2), =√(0.0042+(0.0402-
0.0042)+(0.1622-0.0042)), had an anticipated SD in the
master versus milled group of 0.167 mm; the actual SD
was 0.170 mm. Our results show that the implant position
is the greatest source of error in the milling of an implant
crown. Software (0.004), scanning (0.040), and the milling
process (0.162) and the milled implant scan body (0.170
mm) all contributed error The appreciable variation in
implant location may result from the attachment of the
analog into the milled model.16 The larger variation at the
implant compared with the other groups can be attributed
to the friction inside the socket in which the analog was
placed and to the large changes in curvature on the milled
implant. Friction caused variation in reproducing the
vertical height of the analog between scans and other
studies have confirmed that large changes of curvature
result in larger deviations for 3D scanners.11
This study showed that the milling of an implant
crown is highly accurate and reliable. Following the
implant group (0.014 ±0.170 mm), the anatomic sites
with the greatest amount of variation were the cusps and
occlusal ridge and fossa whose mean volumetric devia-
tion was −0.038 ±0.023 mm and −0.032 ±0.027 mm,
respectively (Table 2) (master versus milled). Negative
deviations will not occur in the oral cavity when resto-
rations are seated,15 and a small negative deviation may
reduce the need for occlusal adjustment when the crown
is seated. Using preset fixed measurement points in
this study did not account for marginal discrepancies
in nonmeasured locations. The anatomic site with the
Koch et al
753
smallest amount of variation was at the interproximal
region at 0.000 ±0.013 mm.
The results of this study can be used to minimize
systematic error in the digital workflow of both intraoral
scanners and the milling process. As this was an in vitro
study, the clinical significance of the accuracy and pre-
cision still remains in question. Future studies should
explore the applications of these findings using an in vivo
randomized control trial.
CONCLUSIONS
Based on the findings of this in vitro study, the following
conclusions were made:
1. Software and scanner variation were negligible.
Variations in the milled models resulting from
software and scanner error exhibited statistical sig-
nificance (P<.001).
2. Cumulative errors were found in the line of work-
flow. Software, scanner, and milling error (0.001,
0.021, and 0.098 mm respectively) were shown to
propagate throughout the digital workflow to the
milled model (0.100 mm). The uncertainty
computed from the use of the variance formula may
be somewhat less than the actual. The variables
considered may not fully represent all the contrib-
utors to error.
3. Direct digitization by means of the intraoral scanner
had less systematic error than physical model
fabrication step with the milling machine from the
intraoral scan. The volumetric deviations may not be
interpreted as the absolute difference of a specific
location of one cast compared with another. How-
ever, the pooled locations describe the reliability of
the milling process as it applies to specific anatomic
locations on the tooth.
REFERENCES
1. Andersson M, Oden A. A new all-ceramic crown. A dense-sintered,
high-purity alumina coping with porcelain. Acta Odontol Scand 1993;51:
59-64.
2. Andersson M, Carlsson L, Persson M, Bergman B. Accuracy of machine
milling and spark erosion with a CAD/CAM system. J Prosthet Dent 1996;76:
187-93.
3. Mormann WH. The evolution of the CEREC system. J Am Dent Assoc
2006;137 Suppl:7S-13S.
4. Ender A, Mehl A. Full arch scans: conventional versus digital impressionsean
in-vitro study. Int J Comput Dent 2011;14:11-21.
5. Luthardt RG, Loos R, Quaas S. Accuracy of intraoral data acquisition in
comparison to the conventional impression. Int J Comput Dent 2005;8:
283-94.
6. Christensen GJ. Impressions are changing: deciding on conventional, digital
or digital plus in-office milling. J Amer Dent Assoc 2009;140:1301-4.
7. Ender A, Mehl A. Accuracy of complete-arch dental impressions: A new
method of measuring trueness and precision. J Prosthet Dent 2013;109:121-8.
8. Flugge TV, Schlager S, Nelson K, Nahles S, Metzger MC. Precision of
intraoral digital dental impressions with iTero and extraoral digitization with
the iTero and a model scanner. Am J Orthod Dentofacial Orthop 2013;144:
471-8.
9. International Organization for Standardization. ISO-5725-1. Accuracy
(trueness and precision) of measurement methods and results d Part 1:
THE JOURNAL OF PROSTHETIC DENTISTRY
754 Volume 115 Issue 6

General principles and definitions. Geneva: ISO; 1994. Available at: 15. Guth JF, Keul C, Stimmelmayr M, Beuer F, Edelhoff D. Accuracy of digital
https://www.iso.org/obp/ui/#iso:std:iso:5725:-1:ed-1:v1:en. Accessed models obtained by direct and indirect data capturing. Clin Oral Investig
May 19, 2015. 2012;17:1201-8.
10. Ziegler M. Digital impression taking with reproducibly high precision. Int J 16. Lee SJ, Betensky RA, Gianneschi GE, Gallucci GO. Accuracy of digital versus
Comput Dent 2009;12:159-63. conventional implant impressions. Clin Oral Implants Res 2015;26:715-9.
11. Rudolph H, Luthardt RG, Walter MH. Computer-aided analysis of the in- 17. Ku H. Notes on the use of propagation of error formulas. J Res Natl Bur
fluence of digitizing and surfacing on the accuracy in dental CAD/CAM Stand 1966;70:263-73.
technology. Comput Biol Med 2007;37:579-87.
12. Persson A, Andersson M, Oden A, Sandborgh-Englund G. A three-dimen-
sional evaluation of a laser scanner and a touch-probe scanner. J Prosthet Corresponding author:
Dent 2006;95:194-200. Dr Sang J Lee
13. Persson AS, Andersson M, Oden A, Sandborgh-Englund G. Computer aided Harvard School of Dental Medicine
analysis of digitized dental stone replicas by dental CAD/CAM technology. Department of Restorative Dentistry and Biomaterials Science
Dent Mater 2008;24:1123-30. 188 Longwood Ave
14. Steinhauser-Andresen S, Detterbeck A, Funk C, Krumm M, Kasperl S, Boston, MA 02115
Holst A, et al. Pilot study on accuracy and dimensional stability of Email: sang_lee@hsdm.harvard.edu
impression materials using industrial CT technology. J Orofac Orthop
2011;72:111-24. Copyright © 2016 by the Editorial Council for The Journal of Prosthetic Dentistry.

Noteworthy Abstracts of the Current Literature

Immediate versus conventional loading of complete-arch implant-supported

prostheses in mandibles with failing dentition: A patient-centered controlled
prospective study

Peñarrocha-Oltra D, Peñarrocha-Diago M, Aloy-Prosper A, Covani U, Peñarrocha M: Int J

Prosthodont 2015;28:499-508

Purpose. The aim of this study was to compare, from the patients’ perspective, immediate and conventional loading
of fixed complete-archprostheses to rehabilitate mandibles with failing dentition.
Materials and methods. This controlled, prospective, nonrandomized study included 36 consecutive patients:
18 treated with conventional loading (control) and 18 with immediate loading (test). Patient general satisfaction
and specific satisfaction with esthetics, chewing, speaking, comfort, self-esteem, ease of cleaning, and treatment
duration were evaluated using 10-cm visual analog scales before treatment and 3 and 12 months after treatment.
Postoperative pain and swelling were monitored daily for 1 week. Statistical analysis was performed applying
Mann-Whitney and Wilcoxon tests (a=.05).
Results. Between baseline and 3 months, satisfaction in the test group increased significantly with the exception
of speech; in the control group, satisfaction increased significantly for esthetics and decreased significantly for speech,
chewing, and comfort, but did not vary for general satisfaction or self-esteem. After 3 months, satisfaction was
significantly higher in the test group with the exception of ease of cleaning. Between 3 and 12 months, satisfaction
improved in both groups but more so in the control group, so that after 12 months there were no differences. The
test group showed lower mean pain, which began after the third day postsurgery. Mean swelling and maximum
pain/swelling did not show significant differences at any point.
Conclusions. Patient satisfaction was reported as significantly higher with immediate loading. However, at the end of
the observation periods, reported functional differences had disappeared. Significant differences were only noted
for postoperative pain after the third day.

Reprinted with permission of Quintessence Publishing.

THE JOURNAL OF PROSTHETIC DENTISTRY Koch et al