Fatigue Behavior of Computer-Aided Design/

Computer-Assisted Manufacture Ceramic Abutments as a

Function of Design and Ceramics Processing
J. Robert Kelly, DDS, MS, DMedSc1/Patchnee Rungruanganunt, DMD, MSD2

Purpose: Zirconia is being widely used, at times apparently by simply copying a metal design into ceramic.
Structurally, ceramics are sensitive to both design and processing (fabrication) details. The aim of this work
was to examine four computer-aided design/computer-assisted manufacture (CAD/CAM) abutments using a
modified International Standards Organization (ISO) implant fatigue protocol to determine performance as
a function of design and processing. Materials and Methods: Two full zirconia and two hybrid (Ti-based)
abutments (n = 12 each) were tested wet at 15 Hz at a variety of loads to failure. Failure probability distributions
were examined at each load, and when found to be the same, data from all loads were combined for lifetime
analysis from accelerated to clinical conditions. Results: Two distinctly different failure modes were found
for both full zirconia and Ti-based abutments. One of these for zirconia has been reported clinically in the
literature, and one for the Ti-based abutments has been reported anecdotally. The ISO protocol modification
in this study forced failures in the abutments; no implant bodies failed. Extrapolated cycles for 10% failure
at 70 N were: full zirconia, Atlantis 2 × 107 and Straumann 3 × 107; and Ti-based, Glidewell 1 × 106
and Nobel 1 × 1021. Under accelerated conditions (200 N), performance differed significantly: Straumann
clearly outperformed Astra (t test, P = .013), and the Glidewell Ti-base abutment also outperformed Atlantis
zirconia at 200 N (Nobel ran-out; t test, P = .035). Conclusion: The modified ISO protocol in this study
produced failures that were seen clinically. The manufacture matters; differences in design and fabrication
that influence performance cannot be discerned clinically. Int J Oral Maxillofac Implants 2016;31:601–609.
doi: 10.11607/jomi.4698

Keywords: CAD/CAM abutments, fatigue, implant abutments, titanium, zirconia

C omputer-aided design/computer-assisted manufac-

ture (CAD/CAM) zirconia abutments, like all ceramic
parts, will have performance lifetimes determined by (1)
design details (eg, creating or avoiding stress concen-
trations) and (2) processing defects (eg, sintering flaws,
residual machining damage, large cubic grains). Recent
examples of very poor ceramic design include a zirconia
implant and a zirconia abutment both obviously designed
for fabrication from metals. A nearly 10% failure rate in
3 years was reported for the Z-Look3 zirconia implant
1Professor, Department of Reconstructive Sciences, University
of Connecticut Health Center, Farmington, Connecticut, USA.
2 Associate Clinical Professor, Department of Reconstructive
Sciences, University of Connecticut Health Center,
Farmington, Connecticut, USA.
Correspondence to: Dr J. Robert Kelly, Department of
Reconstructive Sciences, University of Connecticut Health
Center, 263 Farmington Avenue, Farmington, CT 06030-1615,
USA. Fax: 860-679-1360. Email: kelly@uchc.edu
©2016 by Quintessence Publishing Co Inc.
(Z-Systems).1 In general, these fractured from the root
of the first exposed thread. Light and scanning electron
microscopy images and fracture surface analysis revealed
very sharp thread roots and an apparently sandblasted
surface; the first was expected to concentrate stresses
and the second to weaken the ceramic. In essence, while
probably not a bad design for titanium, these implants
fabricated in ceramic were designed for failure. Next is a
zirconia abutment (Procera Zirconia, Nobel Biocare) that
was designed to be inserted with a titanium screw with
a 45-degree angled portion seating against a 45-degree
angled zirconia seat.2 While likely a good screw seat design
for titanium parts to maximize the retention of clamp-
ing force, this arrangement places high hoop stresses
into the zirconia abutment, almost guaranteeing early
failure. Five abutments fractured within 28 to 188 days.
Quantitative fracture surface analysis of three yielded
calculated failure stresses appropriate for transformation
toughened zirconia2; hence, this was not bad material
but poor use of a good material.
While the excellent systematic review of implant
abutments by Zembic et al3 indicated no difference in
performance at 5 years between titanium and zirconia

The International Journal of Oral & Maxillofacial Implants 601

© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
Kelly/Rungruanganunt
abutments, only 134 of the 2,999 abutments were
ceramic, and these were stock parts. Stock ceramic
parts are conceivably much easier for a manufacturer
to design and quality control.
This project examined commercially available custom
CAD/CAM abutments for failure behavior and failure
origins to gain insights into the role of abutment design
and manufacturing on lifetimes. This was not intended
to be a comparison of commercial products. Commercial
products, all designed for the same implant, are needed
to explore the design and manufacturing flaws inherent
to such parts on the market today. The goal of this study
was to perform clinically relevant mechanical testing
to provide practitioners with evidence for choosing
the designs and materials to protect and benefit their
patients and potentially to provide manufacturers with
feedback regarding processing and design issues. In
support of this, an equally important aim was to validate
a fatigue protocol derived from ISO 14801:2007 (Inter-
national Standards Organization)4 as creating failure
modes seen clinically.
To proceed with the most efficient experimental
design, given that this was virtually novel research
involving expensive parts, the project was completed
in two phases. Data from Phase 1 were used for the
statistical design of Phase 2. In essence, Phase 1 explored
one failure load with an initial six specimens to help
direct further testing with an additional 12 specimens
at other loads.
The overall objectives of the study were:
1. Determine if there were differences in
performance attributable to design or
manufacturing, and if yes, how these differences
translate into a lifetime prediction for the
restoration.
2. If there were no differences in performance,
what were possible differentiating factors
among ceramic abutments related to design or
manufacturing (eg, failure origins) that might
drive further improvements in all systems?
MATERIALS AND METHODS
Abutments
One implant type was used throughout the entire
project: Straumann bone level (4.1 mm, 10 mm; Institute
Straumann). Four commercial custom CAD/CAM abut-
ments were fabricated via a private dental laboratory
(Yankee Dental Arts). With one exception, this laboratory
was careful to submit abutment orders piecemeal rather
than in bulk to obscure that they were identical parts
intended for research and not clinical purposes. Initially,
the project was planned for studying four all-zirconia
602 Volume 31, Number 3, 2016
© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
abutments, but by the time of approval, only two manu-
facturers still offered such abutments for the chosen
implant. By agreement with the Research Committee, ITI
Foundation, the study was changed to examine: (1) two
full zirconia abutments (Straumann Cares; Astra Atlantis,
Dentsply) and (2) two titanium-zirconia (Ti-base) hybrids
(Nobel Biocare; Glidewell Laboratories). Interestingly,
cementation of the zirconia sleeves onto the titanium
tubes differed per manufacturer; Nobel Biocare leaves
this step to be accomplished by the ordering dental
laboratory, and Glidewell completes this step in-house.
For the Nobel Biocare abutments, the zirconia sleeves
and titanium tubes were sandblasted (aluminum oxide,
50 μm; 1 bar on zirconia and 2 bar on titanium), and then
the sleeves were cemented with the Multilink Automix
Implant (Ivoclar Vivadent). Custom zirconia crowns were
fabricated for each abutment type with a hemispheri-
cal occlusal surface, with the center of the hemisphere
being 11 mm from the base (per ISO 14801:2007).
Rotational Play
Rotational play of the Phase 1 abutments was measured
using a purpose-built device (Institute Straumann).
Means for the three measured for each of the six abut-
ments were obtained (degrees, minutes).
Fixturing and Loading
Bases were machined from a woven glass fiber-filled
epoxy commercially available as National Electrical
Manufacturers Association G10 (NEMA G10). This mate-
rial has an elastic modulus of approximately 18 GPa,
serving well as a “bone-like” analog and being within
the range specified within ISO 14801:2007. The implants
were screwed into bases until the coronal extent was
flush with the G10 bases. The abutments were not
adjusted in any way prior to delivery and were delivered
with manufacturer-provided abutment screws torqued
to 35 Ncm (clinical torque wrench, Straumann).
The protocol was an adaptation of ISO 14801:2007
with the 3 mm of simulated bone loss eliminated to
force the highest stresses into the abutments (Ivoclar
has chosen to make the same modification as part of
their in-house abutment fatigue research [Albrecht T,
personal communication, 2015]). This 3 mm was added
to the coping height to maintain the moment arm speci-
fied in ISO 14801:2007. The abutment assemblies were
loaded at 30 degrees, at 15 Hz in water, all in keeping
with ISO 14801:2007. The only other deviation from the
ISO protocol involved loading with a bearing chase to
minimize lateral loading versus the ISO protocol involv-
ing loading with a 50-cm rod. At loads of approximately
300 N, lateral loading is less than 10 N with this arrange-
ment. All abutment-crown heights were measured using
a dental surveyor and a micrometer to assure control of
moment arm length.
 Kelly/Rungruanganunt

Specimens were loaded in servo-electrical fatigue

8  Straumann
machines (Bose EnduraTEC Elf 3300). The loads were Astra Atlantis
sinusoidal from 20 N to the target load. Failure was Glidewell
defined as 0.5-mm displacement beyond the plateau 7 
Nobel Biocare
value achieved during the first 1,000 cycles. Displace-

Degrees (Min)
ment greater than 0.5 mm triggered machine shutdown 6 
capturing the failure cycle. The authors never encoun-
tered a machine shutdown event that did not involve 5 
a failed part. All failed abutments were photographed
with a stereo light microscope (Leica MX95, Leica Micro- 4 
systems), with selected specimens further evaluated by
scanning electron microscopy (SEM; JEOL JSM-5900LV, 3 

St Zirc

As (Zir
JOEL). Failure origins were determined wherever pos-

ra on

No on
tra co

(Z
(Z
(
um ia

irc
irc

be ia-t
Gl ia-t
At nia
an )

on

l B ita
sible. Some specimens were shared with Dr Isabelle

idw ita
lan )
n

ioc niu
tis

ell niu

ar m)
Denry at the University of Iowa for analysis; she has been

quite active in studying clinically failed abutments from
their implant clinics, with special emphasis on looking
for thermal processing problems leading to the forma-
tion of large grains of cubic zirconia.
Phase 1 and Phase 2 Rationale
Phase 1 involved preliminary testing of six abutments
per manufacturer, all at 200 N. This provided initial data
on the failure loads and cycles for designing the larger
Phase 2 program involving an additional 12 abutments
per group. Breaking the project into two phases also
potentially allowed screening for “batch” effects.
Statistical Design and Analysis (Fatigue Data)
Fatigue data were analyzed using a robust, lifetime anal-
ysis package ALTA 7 (Reliasoft). This package is designed
to accommodate the nature of fatigue data, including
choices of overall failure model (general log-linear
applies to implant and abutment failure mechanisms)
and then three probability distribution functions within
the general log-linear model (Weibull, exponential or
lognormal). These distributions can be “best fit” using
a distribution wizard function.
Fatigue data are generated from at least three dif-
ferent loads. So long as the same probability distribu-
tion function holds for all fatigue data at each testing
load, the general log-linear function can write a global
expression for the probability of failure as a function of
load. This global expression is used to extrapolate the
fatigue data from the accelerated testing conditions
(higher than clinical loads) down to expected clinical
loads. Such extrapolation is legitimate so long as the
failure mechanism(s) under accelerated conditions are
the same as under clinical use. Any probability of failure
can be calculated: eg, 90%, 50%, 10%, depending on
the prediction information desired (eg, total product
lifetime, time for 10% of products to fail). Numerous
graphical outputs are available to explore and compare
the lifetimes of tested parts. Confidence intervals are
used for statistical comparisons.
e
m)
Fig 1  Rotational play of Phase 1 abutments prior to fatigue
testing (means, standard deviations).
RESULTS
Control of Lever Arms
The height of abutments with crowns, from the base
to the top of the hemispherical crowns, was carefully
controlled. It was extremely critical to control this variable
since the moment arm and therefore bending stresses
are determined to a large extent by this. Crown-to-base
lengths were 14 mm (± 0.05 to 0.06 mm), so moment
arm tolerances of approximately 0.04% were achieved.
With the crown sphere having a radius of 3 mm, the
target moment arm under ISO 14801 (sphere center to
implant base) of 11 mm was held.
Rotational Play
Mean rotational play (degrees, minutes) and standard
deviations appear in Fig 1.
Failure Modes
Two distinctly different failure modes were found within
each type of abutment (pure zirconia and Ti-base), and
both occurred in parts from each manufacturer. For
zirconia abutments, failure occurred either from within
the implant, likely prior to screw failure (int), or from the
screw seat (ss); examples are shown in Fig 2. For Ti-base
abutments, failures were due to either titanium tube
tearing (at the base of the tube) not involving zirconia
fracture or by a combination of debonding and zirconia
fracture; examples are shown in Fig 3. It was not possible
to determine whether debonding or ceramic fracture
was the initial event. Failure mode distributions are
seen in Table 1. Although internal fracture appears to
be less for Straumann zirconia abutments, this was not
statistically significant (P > .1, chi-square). Titanium tube
tearing was more prevalent for Glidewell Ti-base than
for Nobel Biocare (P < .001, chi-square).

The International Journal of Oral & Maxillofacial Implants 603

© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
Kelly/Rungruanganunt

Fig 2  Two failure modes encountered

with zirconia abutments. (a) An example
of internal failure (int), where abutment
fracture apparently precedes screw fail-
ure and zirconia can create considerable
titanium wear debris (not seen in this
case). (b) An example of failure from the
screw seat.

a b

Fig 3  Examples of the two failure

modes encountered with Ti-base abut-
ments. (a) The predominant mode for
Glidewell abutments involving tearing of
the titanium tube without fracture of the
zirconia sleeve. (b) A mixture of debond-
ing and zirconia sleeve fracture. It is not
known whether debonding or ceramic
fracture is the precipitating event.

a b

Table 1  Distribution of Failure Modes lognormal distributions as a function of load, allowing

calculation of distributions (probability distribution func-
Debond
Screw and Tube
tions) at any load. At 70 N, only Nobel Biocare separates
Internal seat fracture tear Runout from the others.
Atlantis 8 8 – – –
(16) Failure Origins: Clinical Situation and
Straumann 3 13 – – 1 Examples from Previous Work
(17) Optical images were recorded for all abutments, but only
Glidewell – – 4 13 – representative examples can be shown. Selected zirconia
(17) parts were chosen for SEM analysis to explore for failure
Nobel (14) – – 8 5 1 origins and processing defects. Results from this work will
Distributions were similar for Atlantis and Straumann (P > .1, chi-
be presented following a review of examples known from
square) but differed between Glidewell and Nobel Biocare (P < .001, clinical failures and some preparatory fatigue work spon-
chi-square). Examples of each failure mode can be seen in Figs 3 sored at the Fraunhofer Institute (Stuttgart, Germany).
and 4.
Such comparisons help to establish the generality of the
findings of the present study and their clinical validity.
Fatigue Durability Figure 6 shows an Atlantis abutment (clinical photo)
Overall results appear in Table 2 and Fig 4. Testing was obtained from a former resident that failed in less than 2
begun at 200 N for all systems. This was the upper months. Two similar failures were recently documented
limit for Astra and Glidewell and the lower limit for in published case reports.5 The origin of the titanium
Straumann and Nobel Biocare. Further testing in Phase debris will become apparent in later figures.
2 was performed at lower loads for Astra and Glidewell Figure 7 features images that are also from prepara-
and higher loads for Straumann and Nobel Biocare. tory fatigue work sponsored at the Fraunhofer Insti-
Statistical comparisons can therefore be made at 200 N. tute. This abutment failed from a residual subsurface
At 200 N, Straumann Cares clearly outperformed Astra machining crack circled in Fig 7a and seen at higher
Atlantis (t test, P = .013) with nontransformed mean magnification in Fig 7b. It is also apparent that grinding
cycles-to-failure of: Straumann = 2,534,112 cycles; “dust” remained on the screw seat during processing
Astra = 2,421 cycles. The Glidewell Ti-base abutment and had been lightly sintered, not unexpected given
also outperformed the Atlantis zirconia at 200 N (t test, its low density. This finding is consistent with the screw
P = .035). seat being formed prior to final sintering of the part (ie,
Figure 5 shows the failure probability distributions green machining). Indentation cracks in partially sintered
extrapolated from measured data down to “use loads” zirconia have been reported to grow during final sinter-
of 70 N. These are calculated by ALTA 7 by fitting overall ing (Denry I, personal communication, 2015).

604 Volume 31, Number 3, 2016

© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
 Kelly/Rungruanganunt

Table 2  Global Comparison of All Failure Data and Modes

Load
(N) No. Atlantis Mode No. Glidewell Mode No. Straumann Mode No. Nobel Biocare Mode
150 N 15 3,378,941 ss
16 504,865 ss
17 19,667 int
18 267,687 ss
165 N 11 418,136 int
12 158,875 int
13 9,234 ss
14 2,649 ss
180 N 7 3,063,670 int 7 6,660,748 tt
8 3,256 int 12 117,898 tt
9 444 ss 13 95,352 tt
10 4,887 int 15 187,748 tt
16 115,107 tt
17 35,784 tt
190 N 8 1,495,793 tt
10 57,604 tt
11 78,831 tt
14 11,304,311 df
18 50,537 tt
200 N 3 1,877 int 1 111,868 tt 1 2,000,011 int 1 15,000,020 ro
4 3,872 ss 2 21,994 df 2 11,100,109 ss
5 3,578 ss 3 1,450 df 7 4,508,749 ss
6 355 int 4 138,604 tt 8 25,437 ss
5 1,382,673 tt 9 35,703 ss
6 2,760,542 df 10 54,169 ss
18 14,608 ss
215 N 11 15,228 ss
12 15,476,889 ss
13 6,014,813 ss
15 16,015,296 ss
16 46,305 ss
17 4,186 ss
225 N 5 6,493 int 3 25,000,000 df
14 77,800,024 ro 1,967,891
230 N 6 139,144 tt
7 15,153,000 df
8 6,044,001 df
9 5,359,382 df
10 11,495,677 tt
11 53,800,024 ro
245 N 12 20,503,431 df
13 19,408 df
14 61,475 d(f)
15 8,871,234 df
250 N 4 739 int 2 54,939 tt
16 187,204 tt
17 161,562 tt
300 N 3 680 ss
Failure modes: int = internal; ss = screw seat; df = debond and fracture; tt = tube tear; ro = runout.
Glidewell #9 was accidentally broken by the machine.

The International Journal of Oral & Maxillofacial Implants 605

© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
Kelly/Rungruanganunt

Astra Atlantis
Straumann Cares
20  Glidewell
Nobel Biocare
18 

16 

14 
In cycles

12  * r2 = 0.95

10 

8  *
r2 = 0.98 r2 = 0.83
6 

4 
140 160 180 200 220 240 260
Load (N)

Fig 4   Global comparison of ln (cycles-to-failure) at all loads for all abutments. The natural log transformation was used to make sta-
tistical comparisons possible for the three systems tested at 200 N (Nobel Biocare was tested at 200 N but ran out at 15,000,000
cycles). At 200 N, Glidewell outperformed Atlantis (t test, P = .035).

Astra Atlantis
99.0 
Glidewell Straumann Nobel Biocare Use level
Atlantis\Data 1
General Log-linear
Lognormal
70
F = 16 | S = 0
Data points
Use level line
Probability of failure at 70 N (%)

Glidewell\Data 1
General Log-linear
Lognormal
70
F = 16 | S = 0
50.0  Data points
Use level line
Nobel\Data 1
General Log-linear
Lognormal
70
F = 12 | S = 4
Data points
10.0  Use level line
Straumann\Data 1
5.0  General Log-linear
Lognormal
70
F = 14 | S = 2
1.0  Data points
100,000 1.0 E+8 1.0 E+12 1.0 E+16 1.0 E+20 1.0 E+25 Use level line

Cycles to failure (n)

Fig 5   Global comparison of failure probabilities versus cycles extrapolated to a use load of 70 N. Based on 90% confidence inter-
vals (not shown), Nobel Biocare > Straumann = Atlantis = Glidewell.

606 Volume 31, Number 3, 2016

© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
 Kelly/Rungruanganunt

a b c
Fig 6   (a) Failed Atlantis abutment from private practice (Dr Rob Schulman) that survived for ap-
proximately 2 months. (b) SEM view of fracture surface from a similar Atlantis failure from UConn
Graduate Prosthodontics Clinic. This restored a single maxillary central incisor and fractured within
2 weeks of delivery. Note that the wall thickness is only approximately 0.25 mm. (c) Fracture mark-
ings (twist hackle, at arrows) indicate failure from the surface in likely contact with the screw.

Fig 7   Fatigue failure of Atlantis abutment from Fraunhofer In-

stitute. (a) Machined screw seat surface up (machining grooves
visible). Fracture surface is on the bottom with the origin circled.
(b) Higher magnification of origin but looking down on the screw
seat (lower). Failure appears to have involved a residual sub-
surface machining flaw. Grinding dust on the screw seat has
“lightly sintered,” indicating that machining was done in the
green state.
a b

Fig 8   Fracture origin of Atlantis 13 (165 N; 9,234 cycles) from

the screw seat. (a) Actual origin is visible (arrows). (b) A cross
section of the porous, poorly sintered material remaining on the
machined surface to a depth of approximately 5 μm. a b

Fig 9  Atlantis 6 (200 N; 355 cycles).
Screw hole with the same machining po-
rosity layer in the cross section as in Fig
7 (identified by Dr Isabelle Denry, Univer-
sity of Iowa).
Fig 10   Atlantis 3 (200 N; 1,877 cycles).
Internal origin from surface in apparent
rubbing contact with the implant wall.
Apparent titanium shavings at the ori-
gin indicating micromovement following
zirconia fracture before screw failure. Ar-
rows indicate crack origin.
Fig 11   Straumann 1 (200 N; 2,000,011
cycles). Internal origin with flattened tita-
nium shavings indicating that ceramic
fracture preceded screw failure. Arrows
indicate crack origin.

Failure Origins: Representative Examples from
This Work
Figure 8 is Atlantis 13, which failed in 9,234 cycles at
165 N, showing failure from a subsurface machining
groove. In Fig 8b, the poorly sintered grinding dust
layer is visible (as in Fig 7), and in the cross section, it
is seen to have a depth of approximately 5 μm. Figure
9 contains an image of similar porosity found on
the inner-machined surface of Atlantis 6 (not failure
origin), which survived only 355 cycles at 200 N (SEM
observations and images from Dr Isabelle Denry,
University of Iowa). Such defects were not seen in
Straumann specimens.
Apparent titanium shavings are seen in Fig 10 at
the failure origin of Astra Atlantis 3 following 1,877
cycles at 200 N. Another example of titanium debris is
seen in Fig 11 from the Straumann 1 abutment (200 N;
2,000,011 cycles).

The International Journal of Oral & Maxillofacial Implants 607

© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
Kelly/Rungruanganunt
DISCUSSION
This work modified the ISO 14801 protocol in three
main ways. First, the authors did not want implant
body failures, so the 3 mm of simulated bone loss
was not followed, and 3 mm of crown height was
added to keep the same 11-mm moment arm. This
served to concentrate stresses in the abutments
and resulted, as designed, in all but two abutments
failing (Table 2), and there were no implant fail-
ures. This same approach has been independently
adopted at Ivoclar for their implant abutment devel-
opment research (Albrecht T, personal communica-
tion, 2015). Moment arms were then identical for all
systems and specimens tested (analysis of variance
[ANOVA] df = 29, F = 0.238, P = .869). Second,
all testing was done under water since damage
accumulation and crack development in ceramics
are very different under wet versus dry conditions.
Third, a bearing race was used rather than the 50-cm
loading rod to limit lateral forces on the implants
during testing. In the authors’ experience, they have
seen less than 10 N lateral loading at axial loads of
300 N.
Failures in this work termed “internal” seem very
consistent with the four known clinical failures of
zirconia abutments: one at UConn clinics, one pri-
vate practice example, and two published cases.5
The findings of the present study are also reinforced
by those from the Fraunhofer Institute with this
same abutment, with the Atlantis abutments on
Astra implants,6 and with another Morse taper abut-
ment (Cercon Balance, Ankylos-Friadent) published
by Delben et al7; all systems showed very similar
internal failure. The nature of this damage deserves
follow-up research since it appears to be involved
in both clinical and in vitro failures. One immediate
thought is that internal failures may involve some
sort of damage accumulation due to zirconia-titani-
um contact. However, in the present study, even at
high magnification, no microstructural damage was
detected at these internal contact sites.
Secondly, another mechanism for internal failures
could be due to micromovement at the screw seat
allowing contact between the zirconia abutment
wall and the inner implant surface. This rubbing
point would then become a fulcrum extending the
length of the moment arm from the crown-screw
seat length to the crown-fulcrum point. Thirdly, and
perhaps most likely, such a fulcrum point may also
be created due to elastic deformation of the screw
and/or implant given the rather low modulus of
titanium and its alloys, the relatively small cross sec-
tions of the titanium parts, and the close tolerances
between titanium and zirconia surfaces.
608 Volume 31, Number 3, 2016
© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
It is also clear from the present work and that of
the Fraunhofer Institute that internal failures provide
a mechanism for the generation of considerable tita-
nium debris in cases where ceramic fracture precedes
screw failure. When this occurs, the sharp zirconia
fracture edges can shave-off and “chew on” titanium
particles with each loading cycle until screw failure.
This time period can be considerable in clinical cases,
as is apparent in the titanium tattoo case reports.5
Screw seat failures in all the Atlantis parts exam-
ined involved origins on their tensile sides, while
screw seat failures for Straumann parts involved
crushing damage on their compressive side. Both
compressive and tensile damage at screw seats
were also reported by Mitsias et al6,8 again for the
Atlantis/Astra system. In the present work and that
of the Fraunhofer Institute, machining grooves were
prominent on all parts, and both direct and indirect
evidence of residual subsurface cracks related to
machining were found. Direct evidence of a machin-
ing flaw is seen in Fig 6. Machined surfaces of Atlantis
abutments (screw seat and screw hole) had a layer of
poorly sintered material with considerable porosity
to a depth of approximately 5 μm (Figs 8 and 9). This
presumably resulted from the manufacturer leaving a
layer of machining (or grinding) dust on the surface
during final sintering. Such a grinding dust layer
would have a very low packing density inhibiting
sintering. Such defects would be seen as processing
defects and were not found in Straumann abutments.
As yet, the authors know of no clinical examples of
screw seat failures.
For both debonding/fracture and tube tearing,
Glidewell parts failed at much lower loads and short-
er numbers of cycles than did Nobel Biocare parts;
hence, there must be influential differences in design
and/or manufacturing. The initiating event in failure
of the hybrid Ti-based abutments by debonding/
fracture remains unknown. Debonding may be the
most likely first event followed by fracture of the
now-unsupported zirconia sleeve. Anecdotal reports
of debonding are not uncommon from clinical/dental
laboratory practice (Pages F, personal communica-
tion, 2015). Glidewell Laboratories cemented their
zirconia sleeves prior to delivery, whereas Nobel
Biocare required that the authors (or their laboratory)
do this step. Uncontrolled potential variables include
differences in cements, cementation protocols, and
cement thickness.
Failure of the hybrid Ti-based abutments by tube
tearing was an unexpected but predominant mecha-
nism. Including the obvious differences in tube
heights (Glidewell < Nobel Biocare), many other
differences need to be examined, such as radii of
curvature at tube bases, tube wall thicknesses, and

differences in tube alloy compositions (creating differ-
ences in work hardening and yield stresses).
Failure Rate Analysis
ALTA 7 contains some powerful features for the analy-
sis of data from accelerated testing. All systems were
predicted to experience most failures early during use
with rates plateauing in late life. Quite interestingly,
these systems changed their relative position as the
assumed service load was increased from 70 to 150
to 200 N. At 70 N, Straumann and Atlantis are quite
similar, but at a 200 N use load, Straumann and Nobel
Biocare are similar. Glidewell abutments are predicted
to have the highest failure rate at all use loads. One
possible clinical interpretation is that in patients
and locations experiencing light occlusal loading,
Straumann and Atlantis are both indicated, while for
bruxing patients, Straumann and Nobel Biocare are
equivalent choices.
Clinical Meaning
One finding of validity involves the authors creating
a failure mode for all-zirconia abutments that has
been reported clinically. The data of the present study
can be interpreted in many different ways. Extrapo-
lating to a service load of 70 N yields the following
projections of the number of cycles for 10% of the
abutments to fail: Atlantis, 2 × 107; Straumann, 3 ×
107; Glidewell, 1 × 106; and Nobel Biocare, 1 × 1021.
These projections tend to say that the present study
data at 200 N (very unfavorable to Atlantis) should
not be used to predict clinical behavior. It may very
well be that 200 N is simply an unrealistic load and
should be considered too “accelerated.” However, this
is balanced by the fact that both of the failed clinical
specimens were Atlantis and results of a 3-year-study
reported failure of five of 30 Atlantis zirconia abut-
ments, with all being of “internal” origin.9 Depending
on the number of chewing and swallowing cycles
assumed (often approximately 1.5 × 106 per year),
17% failure at 3 years, or approximately 4.5 × 106
cycles, is worse performance (4.4 times less) than
predicted by the present study data extrapolation.
The projection of the present study would be in close
agreement with this clinical finding if the number of
chewing cycles per year was 6.7 × 106 versus 1.5 ×
106. Since this number is notoriously not well estab-
lished in the literature, and does not include swallow-
ing contacts, the 4.4 times discrepancy could easily be
less. Also, given that the present study extrapolation
is for 70 N and average lateral intraoral forces could
be higher (and that decreased cycles are not linear
but logarithmic with increased force), this laboratory
prediction and clinical finding are likely in pretty good
agreement.
© 2016 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY.
NO PART MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER.
Kelly/Rungruanganunt
CONCLUSIONS
The altered ISO 14801 test in this study created failure
only from abutments (no implant failure). Testing created
two failure mechanisms for both zirconia and Ti-base
abutments. One of the all-zirconia failure mechanisms
(“internal”) has been seen clinically. One of the failure
mechanisms from Ti-based abutments (debonding/
fracture) has been reported anecdotally from clinical
cases. The fatigue durability did not track with rotational
freedom.
Based on data from this study, neither all-zirconia
nor Ti-based abutments were shown to be superior. The
manufacturer matters; for both all-zirconia and Ti-based
abutments, parts from different manufacturers, design
and fabrication differences influencing performance,
appeared extremely similar via clinician examination.
Ceramics processing defects were documented for the
Atlantis abutments (these will likely be found for other
manufacturers as well). Design defects/issues were
suggested for all systems.
Further statistical distinctions will require testing
more specimens at load closure to use loads.
ACKNOWLEDGMENTS
Funding was provided by the ITI Foundation, Basel, Switzerland.
The authors reported no conflicts of interest related to this
study.
REFERENCES
  1. Gahlert M, Burtscher D, Grunert I, Kniha H, Steinhauser E. Failure
analysis of fractured zirconia implants. Clin Oral Implants Res
2012;23:287–293.
  2. Aboushelib MN, Salameh Z. Zirconia implant abutment fracture:
Clinical case reports and precautions for use. Int J Prosthodont
2009;22:616–619.
  3. Zembic A, Kim S, Zwahlen M, Kelly JR. Systematic review of the
survival rate and incidence of biologic, technical and esthetic com-
plications of single implant abutments supporting fixed prostheses.
Int J Oral Maxillofac Implants 2014;29:S99–S116.4.
  4. ISO 14801:2007 Dentistry — Implants — Dynamic fatigue test for
endosseous dental implants. International Standards Organization,
Geneva, Switzerland.
  5. Taylor TD, Klotz MW, Lawton RA. Titanium tattooing associated with
zirconia implant abutments: A clinical report of two cases. Int J Oral
Maxillofac Implants 2014;29:958–960.
  6. Mitsias ME, Thompson VP, Pines M, Silva NR. Reliability and
failure modes of two Y-TZP abutment designs. Int J Prosthodont
2015;28:75–78.
  7. Delben JA, Barão VA, Ferreira MB, et al. Influence of abutment-to-
fixture design on reliability and failure mode of all-ceramic crown
systems. Dent Mater 2014;30:408–416.
  8. Mitsias ME, Silva NR, Pines M, Stappert C, Thompson VP. Reliability
and fatigue damage modes of zirconia and titanium abutments. Int
J Prosthodont 2010;23:56–59.
  9. Ferrari M, Tricarico MG, Cagidiaco MC, et al. 3-year randomized
controlled prospective clinical trial on different CAD-CAM implant
abutments. Clin Oral Implant Dent Relat Res 2016. 2016 Mar 14. doi:
10.1111/cid.12418. [Epub ahead of print]
The International Journal of Oral & Maxillofacial Implants 609
Copyright of International Journal of Oral & Maxillofacial Implants is the property of
Quintessence Publishing Company Inc. and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.