- Email: [email protected]
Evaluation of test protocol variables for dental implant fatigue research
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
available at www.sciencedirect.com
journal homepage: www.intl.elsevierhealth.com/journals/dema
Evaluation of test protocol variables for dental implant fatigue research Cornell K. Lee a , Matthias Karl b , J. Robert Kelly a,∗ a
Department of Reconstructive Sciences, University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030-1615, USA b Department of Prosthodontics, University of Erlangen-Nuremberg, Erlangen, Germany
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objectives. This work begins to explore the influence of cycling rate and environment on
Received 5 March 2009
fatigue testing of dental implants according to the ISO protocol 14801.
Received in revised form
Methods. Twenty-four Straumann implants (4.1 mm × 12 mm) were tested up to five mil-
24 June 2009
lion cycles per ISO 14801: loaded at either 2 or 30 Hz in room air at 25 ◦ C or normal saline
Accepted 2 July 2009
at 37 ◦ C (n = 6 per group). Implant displacements/cycle were captured during all testing. Fracture-surface features were examined using scanning electron microscopy (n = 12). Two complimentary methods were developed to estimate fatigue crack growth rates.
Keywords:
Results. Failures (bulk fracture) were found to be bi-modally distributed, either <350,000
Dental implant complications
cycles or >1.5 million cycles at both cycling rates. Following initial crack formation, fatigue
Fatigue testing
crack growth required merely 1100–4200 cycles to failure. Initial crack pop-in was statisti-
Loading frequency
cally more likely under 2 Hz than 30 Hz (2 , p < 0.05) but testing in air and normal saline were
Environmental conditions
equivalent in terms of likelihood of fracture versus runout (2 , p > 0.6). On a microscopic
Scanning electron microscopy
level, fatigue crack growth rates appears to be similar at 2 and 30 Hz, but may be slower in the presence of saline versus dry at 2 Hz. Significance. Implant failure under fatigue conditions involved “classic” damage mechanisms. Failure appears more likely at 2 Hz than 30 Hz for reasons that remain to be elucidated. Saline may enable chemically assisted crack growth involving grain boundaries during the stage of fatigue crack growth, but did not influence likelihood of failure. © 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
With growing experience in implant dentistry it appears that osseointegration is a very predictable process. However, technical complications such as fractures of implant components may limit the lifetime of an implant-supported reconstruction [1]. While reports of bulk implant fracture are relatively infrequent, the literature may be underestimating its prevalence. The studies that are available are mostly limited to 5–7
∗
years of follow-up [2–10] with only a few that have a maximum follow-up beyond 9 years. In a 15-year study, Adell et al. found an implant fracture incidence of 3.5% with most of the fractures occurring after 5 years of clinical function [11]. This may indicate that a 5-year follow-up is inadequate in studying the process of implant fractures. Furthermore, the implants included in these studies mainly involved either fixed complete or fixed partial dentures. It has been reported that mechanical complications occur more frequently in single
Corresponding author. Tel.: +1 860 679 3747; fax: +1 860 679 1370. E-mail address: [email protected] (J.R. Kelly). 0109-5641/$ – see front matter © 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.dental.2009.07.003
1420
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
tooth restorations and more so in the posterior region than the anterior region [12,13]. Additionally, with implants expected to serve for decades, fatigue failure may still emerge as an issue. In vitro implant testing protocols in the literature vary among studies, making it difficult to compare results. Some studies have tested dental implants using monotonic loading [14,15] while others have used cyclic loading [16–18]. Unfortunately, monotonic loading may have very little clinical relevance as mechanical failures in dentistry are more likely related to a long term process with repeated low loads rather than an acute overload [19]. Published cyclic testing protocols have utilized various implant loading angles, frequency of load, application load levels and length of lever arm on the implant. Furthermore, most published studies tested only dry conditions and provide little or no documentation of the failure modes. A standardized implant fatigue testing protocol was developed in 2003 by a panel of industry and academic experts for the Organization for International Standardization (ISO 14801). The ISO recommendations were designed for single, endosteal, transmucosal dental implants tested under “worst case” applications. However, the standard was created based on limited experimental data, including limited guidance indicating that implant fracture mechanisms do or do not differ depending on cycling rate or water immersion. Current guidelines limit testing in wet conditions to 2 Hz carried out until failure or two million cycles. For dry conditions testing is limited to 15 Hz carried out until failure or five million cycles. Testing at 2 Hz would require 12 days while testing at 15 Hz would last only 4 days. This present work begins to explore the influence of certain ISO fatigue protocol variables (cycling rate and environment) to rationalize implant testing procedures and lay some groundwork for later validation of in vitro methods as reproducing clinically relevant failure mechanisms. Understanding the influence of cycling rate and environment is also important in rationally expediting implant testing procedures and may provide important guidance for future revisions of ISO 14801.
2.
Materials and methods
The test set-up and conditions were performed according to the ISO 14801 protocol (Fig. 1) with the exception that a bearing race was used to limit lateral loading on the implants versus a 50 mm long loading member attached to a universal joint. Testing involved 4.1 mm × 12 mm, solid screw, standard plus implants with corresponding 5.5 mm tall solid abutments (Straumann AG, Basel, Switzerland). Wet and dry testing at two cycling levels of 2 and 30 Hz was performed with six implants in each experimental group. A testing frequency of 30 Hz instead of 15 Hz (highest allowed under ISO 14801) was chosen to exaggerate any frequency effect. Testing for all groups was carried out until failure or five million cycles. Failure was defined as an implant displacement of 0.5 mm beyond baseline cyclic values. Testing in a wet environment was performed at 37 ◦ C in 0.9% saline (Fischer Scientific, Fairlawn, NJ). Dry testing was performed at room temperature.
Fig. 1 – Implant specimen loaded under conditions fulfilling intention of ISO 14801 (note: strain gauge shown not used in this study and base is aluminum versus G-10).
The dental implants were fixed in an epoxy resin-glass fiber composite (NEMA Grade G-10 rod, Piedmont Plastics, Charlotte, NC). This embedment material has an appropriate elastic modulus for a bone analog material (app. 20 GPa), is easily machined and is sufficiently tough for cyclic testing. The rod stock was sectioned into blocks and 12 mm deep channels were prepared in the center of the block using a surgical 3.5 mm diameter twist drill (Straumann AG) attached to an engineering lathe. The corresponding 4.1 mm tapping drill (Straumann AG) was then used to a depth of 9 mm. An external micrometer (MK II, Fowler Company, Newton, MA) was used to ensure that the implants were mounted with a simulated bone loss of 3 mm. The abutments were attached to the implant with a torque of 35 N cm using the implant manufacturer’s ratchet (Straumann AG). Cylindrical blocks (5 mm × 12 mm) were cast in a high noble alloy (JP-1, Jensen Industries Inc., North Haven, CT) to fit the abutment and implant. The cylinder was then milled using a computer numerically controlled machine to create a hemispherical crown with a radius of 3 mm and a distance of 11 mm between the simulated bone level and the center of the sphere of the crown (creating the moment arm required under ISO 14801). This casting was then scanned (Cercon, Dentsply Ceramco, Burlington, NJ) and two replica crowns were created in zirconia by green machining and sintering. The crowns were cemented on the abutment using polycarbonate cement (Duco Cement, Devcon, Riviera Beach, FL) and reused in subsequent tests. Uniaxial sinusoidal cyclic loading was performed under load control between 20 and 420 N (ElectroForce 3300 Instrument and Win Test Software, Bose Corporation ElectroForce Systems Group, Eden Prairie, MN, USA). The maximum load was chosen from data derived from pilot studies, hoping to create failures in all implants using the lowest load possible. Implant displacement/cycle was monitored and recorded during all testing. Failure was defined as 0.5 mm displacement beyond that established during initial loading. Fracture-surface features were examined and photographed with scanning electron microscopy (JSM-5600LV, Jeol Ltd., Tokyo, Japan). Under the SEM, the origin of the crack
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
1421
was identified and images were obtained along the path of crack propagation. The coordinates of each image in relation to the origin were recorded to calculate the distance from the origin. For the images with well-defined fatigue striations, 10 representative distances between striations were measured. This increment crack growth per cycle (da/dc) as a function of distance from the origin was curve-fit to choose a reasonable mathematical expression that could empirically describe data from all implants so examined (TableCurve 2D, SYSTAT Software, Richmond, CA). Integrating the inverse of this function (dc/da; cycles/increment) provided an estimate of the number of total cycles needed for fatigue growth to the failure point. A second method was used to determine the number of cycles during the fatigue phase of crack growth by examining spreadsheet data of implant displacement per cycle. An increase in displacement of the implant indicated the onset of fatigue. By reviewing the displacement data of the implant during cycling, the point at which fatigue occurred was roughly identified. The subsequent data was used to calculate and estimate the number of cycles for fatigue failure.
3.
Results
A total of 24 implants were tested. They came from two different lots and were evenly distributed between the four testing conditions: 2 Hz dry; 2 Hz wet; 30 Hz dry and, 30 Hz wet. All together 12 implants fractured. In the dry environment 7/12 implants failed compared to 6/12 in the wet environment. In terms of loading frequency, 4/12 implants tested at 30 Hz fractured compared to 9/12 for those tested at 2 Hz. The influence of wet versus dry testing conditions on implant failure was not statistically significant (2 = 0.5, df = 1, p = 0.682) nor was the influence of lot (2 = 0.333, p > 0.5). However, analysis suggested that implants were more likely to fracture than runout when tested at 2 Hz versus 30 Hz (2 = 4.196, df = 1, p = 0.041). When implants failed by bulk fracture, the crack initiated on the tensile side of the specimen. Failure initiated in the root of the first prominent thread, generally located just above the base, but on occasion was just below the level of the base due to thread positioning. Crack development then occurred preferentially in the direction away from the bulk of the thread (i.e. into the thinnest implant cross-section). Implants were not cycled to complete separation of the parts; final separation was done using a monotonic load. For the seven specimens that failed at 2 Hz, five failed in less than 300,000 cycles and two failed in more than 1.5 million cycles, as can be seen in the displacement/cycle graph in Fig. 2A. The majority of the cycles are represented by the relatively flat portion of each data set. The vertical spike at the end of each data set represents increasing implant displacement due to fatigue crack growth the catastrophic failure that occurs in just a few thousand cycles. For the four specimens that failed at 30 Hz, two failed in less than 350,000 cycles while the other two failed after more than two million cycles (Fig. 2B). Failure origins were easily identified at 70× magnification by fracture-surface features having directionality along the crack path converging onto a single point (reminiscent of twist
Fig. 2 – Measurements of implant displacement during cyclic testing for specimens in wet and dry conditions tested at 420 N (W-wet, D-dry). (A) 2 Hz; (B) 30 Hz.
hackle formed under brittle failure). This initial crack generally arrested approximately 225 m into the implant where fatigue growth began, characterized by fatigue striations representing incremental crack growth per cycle (da/dc). Fatigue crack growth occurred from approximately 225 m to around 1400 m into the implant by which time a visible crack was present. An interesting and consistent difference was noted in surface topography of the fatigue portion between wet and dry samples (SEM images). The overall impression is of a flatter, planar fractured surface for the dry specimens (Fig. 3A) compared to a rougher, more convoluted surface for the wet specimens (Fig. 3B). These differences in surface features become more obvious farther away from the origin. Three distinct stages of failure were revealed at higher magnification: (1) crack pop-in and arrest, (2) fatigue crack growth and (3) catastrophic or ductile failure. Fig. 4 represents a montage of SEM images for a failed implant tested at 30 Hz in dry conditions. A high magnification view of the fracture surface within the crack pop-in region is shown in Fig. 5A. This fracture surface is relatively smooth and no fatigue striations are evident. Stage 2 crack growth exhibits crack growth/arrest marks known as fatigue striations, an example of which is shown in Fig. 5B. Each striation is the result of one load cycle. The general direction of crack propagation is from the bot-
1422
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
Fig. 4 – A montage of SEM images for an implant tested at 30 Hz under dry conditions. The dot indicates the origin. Boxes indicate sites for photomicrographs in Fig. 5. (A) Represents the crack pop-in region, (B) is within the fatigue crack growth region and (C) within the ductile failure region.
Fig. 3 – Fracture surface of 2 Hz specimens 1100 m away from the origin. (A) Dry specimen fracture surfaces were relatively planar due to transgranular fracture path. (B) Wet specimen surfaces were considerably rougher indicating an intergranular fracture path involving grain boundaries.
tom of the image to the top and as expected, the majority of the fatigue striations are aligned perpendicular to this. Ductile failure occurred when the fractured implant specimen was manually detached. Cup-like depressions, also known as dimples, are evident and are representative of this mode of fracture (Fig. 5C).
Fatigue striations leave a record of the incremental crack growth per cycle (da/dc). These steps became larger with distance away from the origin. Increment crack growth per cycle as a function of distance from the origin was curve-fit to choose a reasonable mathematical expression: y = a + b(ln x)
2
(1)
where y is the crack growth increment per cycle (m), x is distance from the failure origin (m) and a and b are fitted constants. This expression was chosen as being the simplest relationship that was well-fit to all data sets between 225 and 1400 m for both 2 and 30 Hz conditions (r2 between 0.52 and 0.77). Curve-fit graphs for all four conditions indicated that fatigue striations begin to appear approximately 225 m from the origin, consistent with the SEM observations. From approximately 750 to 1500 m from the origin, curve-fits
Table 1 – Calculation of the duration and percentage of cycles for fatigue failure based on incremental crack growth measurements on failed implants and Eq. (2) (D = dry, W = wet). Specimen 2.2D 2.4D 2.1W 2.3W 30.10D 30.3W a b
Total cycles to failure 229,364 288,619 214,672 79,926 2,297,637 3,918,266
Approx. cycles for fatigue crack growtha 1891 2851 4216 2326 1442 1442
Approx. time during fatigue crack growth (min) 15 24 35 19 1 1
Actual stepwise crack growth (i.e. fatigue) was found to be limited to these numbers of cycles. Lifetime represents the total number of cycles to failure.
Percentage of cycles for fatigue crack growth (% lifetime)b 0.8 1.0 2 2.9 0.1 0
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
1423
Fig. 6 – Incremental crack growth per cycle as a function of distance from the origin for a pair of 2 Hz specimens: (A) 2 Hz, wet; and, (B) 2 Hz, dry. Dots are measured data; inner line is the fitted curve to Eq. (1); outer lines are 95% confidence boundaries. Cracks are seen to be growing more slowly in the presence of saline.
fatigue crack growth:
Fig. 5 – Higher magnifications of fracture surfaces in Fig. 4: (A) crack pop-in stage of failure corresponding to location (A) in Fig. 4; (B) fatigue crack growth region corresponding to location (B) in Fig. 4; and, (C) ductile failure corresponding to location (C) in Fig. 4.
and 95% confidence intervals for incremental growth differ between 2 Hz wet (Fig. 6A) and 2 Hz dry (Fig. 6B). The distance between fatigue striations for 2 Hz wet become less than that for 2 Hz dry, suggesting a diminished incremental crack growth per cycle in the presence of the saline solution (i.e. slower crack growth). An estimate of the number of cycles needed for fatigue growth to the pre-defined failure point was determined by integrating the inverse of Eq. (1) between 225 and 1400 m to obtain an estimate of the total number of cycles representing
1400
225
1 y
(2)
It appears that fatigue crack growth occurred over approximately 1400–4200 cycles (Table 1). Compared to the total number of cycles to failure, greater than 97% of the lifetime of the implant was used to create an initial crack. The second method, of scanning the implant displacement per cycle data to estimate the number of cycles devoted to fatigue failure yielded similar results (Table 2). Fatigue crack growth occurred over approximately 1000–3600 cycles. This represents less than 1% of the lifetime of the implant specimens.
4.
Discussion
One observation from this work is that cycles to failure at one stress level are bi-modally distributed. This bi-modal distri-
1424
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
Table 2 – Calculation of the duration and percentage of cycles for fatigue failure based on load displacement data on failed implants (D = dry, W = wet). Specimen 2.2D 2.4D 2.1W 2.3W 30.10D 30.3W
Total cycles to failure 229,364 288,619 214,672 79,926 2,297,637 3,918,266
Approx. cycles for fatigue crack growth
Approx. time during fatigue crack growth (min)
1620 1140 1590 Data not available 3600 Not detectable
bution suggests that the implant population may contain two distinct groups, for example a certain percentage containing a major residual machining flaw and the remainder free of that specific defect. Alternatively the bi-modal distribution may be related to characteristics of the alloy and represent an outcome of more complex origins inherent to that particular implant under low-load cyclic testing. In either case larger numbers of implants likely need to be tested in order to more fully explore the full failure distributions for each subset. The need for sophisticated experimental designs and larger numbers of specimens may complicate future studies hoping to make statistical comparisons of clinical interest; e.g., comparing seemingly similar implants from two different manufacturers. Comparing the 2 Hz wet versus 2 Hz dry fracture surfaces, the wet specimens appear to have a smaller crack step per cycle. For wet specimens it appears that more crack growth was intergranular than transgranular and that secondary cracking events associated with grain boundaries were more frequent than for dry specimens. Perhaps the saline solution contributed to a chemically assisted corrosion process during fracture, allowing the applied energy to be dissipated into creating the lateral cracks in multiple planes; consequently, the advancement of the crack front was retarded. Another likely factor is a decreased stress intensity associated with cracks turning away from the fracture plane to follow grain boundaries. It appears from this work that initial crack pop-in was more likely to happen at 2 Hz than at 30 Hz. This finding is counter to previous work suggesting that higher stresses would develop in the root of threads as the “effective” elastic modulus of the base increases at 30 Hz versus 2 Hz (unpublished research from our laboratory dealt with in a companion paper [20]). That the base becomes stiffer with higher cycling rates is seen in a shift to smaller implant displacements, likely due to viscoelastic behavior of the filled epoxy base material. That this leads to higher stresses in the root of implant threads is a finding of finite element analysis. Another phenomenon may be competing with higher stresses under 30 Hz testing. One potential candidate would be “strain rate sensitivity”; under lower strain rates (2 Hz) there may be damage accumulation mechanisms that are not favored to occur at higher strain rates (30 Hz). Future work may involve direct measurement of strain rates near the implant surface at risk (as is the topic of a companion paper [20]). There has been minimal research on understanding implant fatigue, with studies mainly limited to case reports. Velásquez-Plata et al. [21] presented a clinically failed implant,
Percentage of cycles for fatigue crack growth (% lifetime)
14 10 13
0.7 0.4 0.7
2
0.2
claiming that SEM images revealed fatigue striations. However, there is a lack of peak and valleys that are characteristic of fatigue striations. The uniformly spaced lines that they describe appear more consistent with machining grooves. Piattelli et al. reported on three clinically fractured implants [22]. Their features described as fatigue striations appears more consistent with those obtained in our 2 Hz wet specimens. Morgan et al. examined clinically fractured implants and compared them to laboratory fractured implants [23]. Their images demonstrated fatigue striation spacing from 0.1 to 1.0 m consistent with our measurements. They did not confirm our observation that the fatigue crack front changed directions on a local scale likely involving crystallographic grain orientations. All of our observations involving grain orientations and grain boundaries require further validation via analysis of the microstructure of the implants tested.
5.
Conclusions
Preliminary analysis suggests that testing in air and normal saline are equivalent in terms of likelihood of fracture versus runout. Implant fatigue failure involved three distinct stages (all of which can be visualized by SEM along with some quantitative measures): (1) brittle crack pop-in and arrest, (2) fatigue crack growth, and (3) final ductile failure. Failures were found to be bi-modally distributed, either (1) <350,000 cycles or (2) >1.5 million cycles, likely limiting future comparative analysis to subgroups representing different failure mechanisms. On a microscopic level, fatigue crack growth rates appeared to be similar under 2 and 30 Hz testing, but may be different for wet and dry conditions at 2 Hz. The failure process occurred over a very limited number of cycles; the majority of the cycles were used to pop-in the initial crack with catastrophic failure then occurring after just a few thousand cycles of fatigue crack growth. Much work has gone into studying the fracture surfaces of failed implants by SEM. Continuations of this effort may lead to the development of powerful tools for understanding whether different test protocols and different laboratories are producing failure by similar mechanisms. Eventually such tools may be useful in examining clinically failed specimens for the validation of laboratory tests.
Acknowledgements This project was supported by a grant from the ITI Foundation for the Promotion of Oral Implantology, Switzerland.
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
references
[1] Taylor TD. Prosthodontic problems and limitations associated with osseointegration. J Prosthet Dent 1998;79:74–8. [2] Zarb GA, Schmitt A. The longitudinal clinical effectiveness of osseointegrated dental implants: the Toronto study. Part III. Problems and complications encountered. J Prosthet Dent 1990;64:185–94. [3] Tolman DE, Laney WR. Tissue-integrated dental prosthesis complications. Int J Oral Maxillofac Implants 1992;7:477– 84. [4] Naert I, Quirynen M, van Steenberghe D, Darius P. A study of 589 consecutive implants supporting complete fixed prostheses. Part II. Prosthetic aspects. J Prosthet Dent 1992;68:949–56. [5] Pylant T, Triplett RG, Key MC, Brunsvold MA. A retrospective evaluation of endossesous titanium implants in the partially edentulous patient. Int J Oral Maxillofac Implants 1992;7:195–202. [6] Gunne J, Jemt T, Linden B. Implant treatment in partially edentulous patients: a report on prostheses after 3 years. Int J Prosthodont 1994;7:143–8. [7] Lekholm U, Gunne J, Henry P, Higuchi K, Lindén U, Bergström C, van Steenberghe D. Survival of the Branemark implant in partially edentulous jaws: a 10-year prospective multicenter study. Int J Oral Maxillofac Implants 1999;14:639–45. [8] Henry PJ, Laney WR, Jemt T, Harris D, Krogh PH, Polizzi G, Zarb GA, Herrmann I. Osseointegrated implants for single tooth replacement: a prospectve 5-year multicenter study. Int J Oral Maxillofac Implants 1996;11:450–5. [9] Balshi TJ. An analysis and management of fractured implants: a clinical report. Int J Oral Maxillofac Implants 1996;11:660–6. [10] Allen PF, McMillan AS, Smith DG. Complications and maintenance requirements of implant-supported prostheses provided in a UK dental hospital. Br Dent J 1997;182:298–302. [11] Adell R, Lekholm U, Rockler B, Brånemark PI. A 15-year study of osseointegrated implants in the treatment of the edentulous jaw. Int J Oral Surg 1981;10:387–416.
1425
[12] Goodacre CJ, Kan JY, Rungcharassaeng K. Clinical complications of osseointegrated implants. J Prosthet Dent 1999;81:537–52. [13] Eckert SE, Meraw SJ, Cal E, Ow RK. Analysis of incidence and associated fractures with fractured implants; a retrospective study. Int J Oral Maxillofac Implants 2000;15:662–7. [14] Norton MR. An in vitro evaluation of the strength of an internal conical interface compared to a butt joint interface in implants design. Clin Oral Implants Res 1997;8:290–8. [15] Möllersten L, Lockowandt P, Linden L. Comparison of strength and failure mode of seven implant systems: an in vitro test. J Prosthet Dent 1997;78:582–91. [16] Khraisat A, Stegaroiu R, Nomura S, Miyakawa O. Fatigue resistance of two implant/abutment joint designs. J Prosthet Dent 2002;88:604–10. [17] Perriard J, Wiskott WA, Mellal A, Scherrer SS, Botsis J, Belser UC. Fatigue resistance of ITI implant-abutment connectors—a comparison of the standard cone with a novel internally keyed design. Clin Oral Implants Res 2002;13:542–9. ˘ H, S¸ahin S. Dynamic [18] C¸ehreli MC, Akc¸a K, I˙plikc¸ioglu fatigue resistance of implant-abutment junction in an internally notched morse-taper oral implant: influence of abutment design. Clin Oral Implants Res 2004;15:459–65. [19] Wiskott HW, Nicholls JI, Belser UC. Stress fatigue: basic principles and prosthodontic implications. Int J Prosthodont 1995;8:105–16. [20] Karl M, Kelly JR. Influence of loading frequency on implant failure under cyclic fatigue conditions. Dent Mater 2009;25:1426–32. [21] Velásquez-Plata D, Lutonsky J, Oshida Y, Jones R. A close-up look at an implant fracture: a case report. Int J Periodontics Restorative Dent 2002;22:482–91. [22] Piatelli A, Piatelli M, Scarano A, Montesani L. Light and scanning electron microscopic report of four fractured implants. Int J Oral Maxillofac Implants 1998;13:561–4. [23] Morgan MJ, James DF, Pilliar RM. Fracture of the fixture component of an osseointegrated implant. Int J Oral Maxillofac Implants 1993;8:409–14.
available at www.sciencedirect.com
journal homepage: www.intl.elsevierhealth.com/journals/dema
Evaluation of test protocol variables for dental implant fatigue research Cornell K. Lee a , Matthias Karl b , J. Robert Kelly a,∗ a
Department of Reconstructive Sciences, University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030-1615, USA b Department of Prosthodontics, University of Erlangen-Nuremberg, Erlangen, Germany
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objectives. This work begins to explore the influence of cycling rate and environment on
Received 5 March 2009
fatigue testing of dental implants according to the ISO protocol 14801.
Received in revised form
Methods. Twenty-four Straumann implants (4.1 mm × 12 mm) were tested up to five mil-
24 June 2009
lion cycles per ISO 14801: loaded at either 2 or 30 Hz in room air at 25 ◦ C or normal saline
Accepted 2 July 2009
at 37 ◦ C (n = 6 per group). Implant displacements/cycle were captured during all testing. Fracture-surface features were examined using scanning electron microscopy (n = 12). Two complimentary methods were developed to estimate fatigue crack growth rates.
Keywords:
Results. Failures (bulk fracture) were found to be bi-modally distributed, either <350,000
Dental implant complications
cycles or >1.5 million cycles at both cycling rates. Following initial crack formation, fatigue
Fatigue testing
crack growth required merely 1100–4200 cycles to failure. Initial crack pop-in was statisti-
Loading frequency
cally more likely under 2 Hz than 30 Hz (2 , p < 0.05) but testing in air and normal saline were
Environmental conditions
equivalent in terms of likelihood of fracture versus runout (2 , p > 0.6). On a microscopic
Scanning electron microscopy
level, fatigue crack growth rates appears to be similar at 2 and 30 Hz, but may be slower in the presence of saline versus dry at 2 Hz. Significance. Implant failure under fatigue conditions involved “classic” damage mechanisms. Failure appears more likely at 2 Hz than 30 Hz for reasons that remain to be elucidated. Saline may enable chemically assisted crack growth involving grain boundaries during the stage of fatigue crack growth, but did not influence likelihood of failure. © 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
With growing experience in implant dentistry it appears that osseointegration is a very predictable process. However, technical complications such as fractures of implant components may limit the lifetime of an implant-supported reconstruction [1]. While reports of bulk implant fracture are relatively infrequent, the literature may be underestimating its prevalence. The studies that are available are mostly limited to 5–7
∗
years of follow-up [2–10] with only a few that have a maximum follow-up beyond 9 years. In a 15-year study, Adell et al. found an implant fracture incidence of 3.5% with most of the fractures occurring after 5 years of clinical function [11]. This may indicate that a 5-year follow-up is inadequate in studying the process of implant fractures. Furthermore, the implants included in these studies mainly involved either fixed complete or fixed partial dentures. It has been reported that mechanical complications occur more frequently in single
Corresponding author. Tel.: +1 860 679 3747; fax: +1 860 679 1370. E-mail address: [email protected] (J.R. Kelly). 0109-5641/$ – see front matter © 2009 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.dental.2009.07.003
1420
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
tooth restorations and more so in the posterior region than the anterior region [12,13]. Additionally, with implants expected to serve for decades, fatigue failure may still emerge as an issue. In vitro implant testing protocols in the literature vary among studies, making it difficult to compare results. Some studies have tested dental implants using monotonic loading [14,15] while others have used cyclic loading [16–18]. Unfortunately, monotonic loading may have very little clinical relevance as mechanical failures in dentistry are more likely related to a long term process with repeated low loads rather than an acute overload [19]. Published cyclic testing protocols have utilized various implant loading angles, frequency of load, application load levels and length of lever arm on the implant. Furthermore, most published studies tested only dry conditions and provide little or no documentation of the failure modes. A standardized implant fatigue testing protocol was developed in 2003 by a panel of industry and academic experts for the Organization for International Standardization (ISO 14801). The ISO recommendations were designed for single, endosteal, transmucosal dental implants tested under “worst case” applications. However, the standard was created based on limited experimental data, including limited guidance indicating that implant fracture mechanisms do or do not differ depending on cycling rate or water immersion. Current guidelines limit testing in wet conditions to 2 Hz carried out until failure or two million cycles. For dry conditions testing is limited to 15 Hz carried out until failure or five million cycles. Testing at 2 Hz would require 12 days while testing at 15 Hz would last only 4 days. This present work begins to explore the influence of certain ISO fatigue protocol variables (cycling rate and environment) to rationalize implant testing procedures and lay some groundwork for later validation of in vitro methods as reproducing clinically relevant failure mechanisms. Understanding the influence of cycling rate and environment is also important in rationally expediting implant testing procedures and may provide important guidance for future revisions of ISO 14801.
2.
Materials and methods
The test set-up and conditions were performed according to the ISO 14801 protocol (Fig. 1) with the exception that a bearing race was used to limit lateral loading on the implants versus a 50 mm long loading member attached to a universal joint. Testing involved 4.1 mm × 12 mm, solid screw, standard plus implants with corresponding 5.5 mm tall solid abutments (Straumann AG, Basel, Switzerland). Wet and dry testing at two cycling levels of 2 and 30 Hz was performed with six implants in each experimental group. A testing frequency of 30 Hz instead of 15 Hz (highest allowed under ISO 14801) was chosen to exaggerate any frequency effect. Testing for all groups was carried out until failure or five million cycles. Failure was defined as an implant displacement of 0.5 mm beyond baseline cyclic values. Testing in a wet environment was performed at 37 ◦ C in 0.9% saline (Fischer Scientific, Fairlawn, NJ). Dry testing was performed at room temperature.
Fig. 1 – Implant specimen loaded under conditions fulfilling intention of ISO 14801 (note: strain gauge shown not used in this study and base is aluminum versus G-10).
The dental implants were fixed in an epoxy resin-glass fiber composite (NEMA Grade G-10 rod, Piedmont Plastics, Charlotte, NC). This embedment material has an appropriate elastic modulus for a bone analog material (app. 20 GPa), is easily machined and is sufficiently tough for cyclic testing. The rod stock was sectioned into blocks and 12 mm deep channels were prepared in the center of the block using a surgical 3.5 mm diameter twist drill (Straumann AG) attached to an engineering lathe. The corresponding 4.1 mm tapping drill (Straumann AG) was then used to a depth of 9 mm. An external micrometer (MK II, Fowler Company, Newton, MA) was used to ensure that the implants were mounted with a simulated bone loss of 3 mm. The abutments were attached to the implant with a torque of 35 N cm using the implant manufacturer’s ratchet (Straumann AG). Cylindrical blocks (5 mm × 12 mm) were cast in a high noble alloy (JP-1, Jensen Industries Inc., North Haven, CT) to fit the abutment and implant. The cylinder was then milled using a computer numerically controlled machine to create a hemispherical crown with a radius of 3 mm and a distance of 11 mm between the simulated bone level and the center of the sphere of the crown (creating the moment arm required under ISO 14801). This casting was then scanned (Cercon, Dentsply Ceramco, Burlington, NJ) and two replica crowns were created in zirconia by green machining and sintering. The crowns were cemented on the abutment using polycarbonate cement (Duco Cement, Devcon, Riviera Beach, FL) and reused in subsequent tests. Uniaxial sinusoidal cyclic loading was performed under load control between 20 and 420 N (ElectroForce 3300 Instrument and Win Test Software, Bose Corporation ElectroForce Systems Group, Eden Prairie, MN, USA). The maximum load was chosen from data derived from pilot studies, hoping to create failures in all implants using the lowest load possible. Implant displacement/cycle was monitored and recorded during all testing. Failure was defined as 0.5 mm displacement beyond that established during initial loading. Fracture-surface features were examined and photographed with scanning electron microscopy (JSM-5600LV, Jeol Ltd., Tokyo, Japan). Under the SEM, the origin of the crack
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
1421
was identified and images were obtained along the path of crack propagation. The coordinates of each image in relation to the origin were recorded to calculate the distance from the origin. For the images with well-defined fatigue striations, 10 representative distances between striations were measured. This increment crack growth per cycle (da/dc) as a function of distance from the origin was curve-fit to choose a reasonable mathematical expression that could empirically describe data from all implants so examined (TableCurve 2D, SYSTAT Software, Richmond, CA). Integrating the inverse of this function (dc/da; cycles/increment) provided an estimate of the number of total cycles needed for fatigue growth to the failure point. A second method was used to determine the number of cycles during the fatigue phase of crack growth by examining spreadsheet data of implant displacement per cycle. An increase in displacement of the implant indicated the onset of fatigue. By reviewing the displacement data of the implant during cycling, the point at which fatigue occurred was roughly identified. The subsequent data was used to calculate and estimate the number of cycles for fatigue failure.
3.
Results
A total of 24 implants were tested. They came from two different lots and were evenly distributed between the four testing conditions: 2 Hz dry; 2 Hz wet; 30 Hz dry and, 30 Hz wet. All together 12 implants fractured. In the dry environment 7/12 implants failed compared to 6/12 in the wet environment. In terms of loading frequency, 4/12 implants tested at 30 Hz fractured compared to 9/12 for those tested at 2 Hz. The influence of wet versus dry testing conditions on implant failure was not statistically significant (2 = 0.5, df = 1, p = 0.682) nor was the influence of lot (2 = 0.333, p > 0.5). However, analysis suggested that implants were more likely to fracture than runout when tested at 2 Hz versus 30 Hz (2 = 4.196, df = 1, p = 0.041). When implants failed by bulk fracture, the crack initiated on the tensile side of the specimen. Failure initiated in the root of the first prominent thread, generally located just above the base, but on occasion was just below the level of the base due to thread positioning. Crack development then occurred preferentially in the direction away from the bulk of the thread (i.e. into the thinnest implant cross-section). Implants were not cycled to complete separation of the parts; final separation was done using a monotonic load. For the seven specimens that failed at 2 Hz, five failed in less than 300,000 cycles and two failed in more than 1.5 million cycles, as can be seen in the displacement/cycle graph in Fig. 2A. The majority of the cycles are represented by the relatively flat portion of each data set. The vertical spike at the end of each data set represents increasing implant displacement due to fatigue crack growth the catastrophic failure that occurs in just a few thousand cycles. For the four specimens that failed at 30 Hz, two failed in less than 350,000 cycles while the other two failed after more than two million cycles (Fig. 2B). Failure origins were easily identified at 70× magnification by fracture-surface features having directionality along the crack path converging onto a single point (reminiscent of twist
Fig. 2 – Measurements of implant displacement during cyclic testing for specimens in wet and dry conditions tested at 420 N (W-wet, D-dry). (A) 2 Hz; (B) 30 Hz.
hackle formed under brittle failure). This initial crack generally arrested approximately 225 m into the implant where fatigue growth began, characterized by fatigue striations representing incremental crack growth per cycle (da/dc). Fatigue crack growth occurred from approximately 225 m to around 1400 m into the implant by which time a visible crack was present. An interesting and consistent difference was noted in surface topography of the fatigue portion between wet and dry samples (SEM images). The overall impression is of a flatter, planar fractured surface for the dry specimens (Fig. 3A) compared to a rougher, more convoluted surface for the wet specimens (Fig. 3B). These differences in surface features become more obvious farther away from the origin. Three distinct stages of failure were revealed at higher magnification: (1) crack pop-in and arrest, (2) fatigue crack growth and (3) catastrophic or ductile failure. Fig. 4 represents a montage of SEM images for a failed implant tested at 30 Hz in dry conditions. A high magnification view of the fracture surface within the crack pop-in region is shown in Fig. 5A. This fracture surface is relatively smooth and no fatigue striations are evident. Stage 2 crack growth exhibits crack growth/arrest marks known as fatigue striations, an example of which is shown in Fig. 5B. Each striation is the result of one load cycle. The general direction of crack propagation is from the bot-
1422
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
Fig. 4 – A montage of SEM images for an implant tested at 30 Hz under dry conditions. The dot indicates the origin. Boxes indicate sites for photomicrographs in Fig. 5. (A) Represents the crack pop-in region, (B) is within the fatigue crack growth region and (C) within the ductile failure region.
Fig. 3 – Fracture surface of 2 Hz specimens 1100 m away from the origin. (A) Dry specimen fracture surfaces were relatively planar due to transgranular fracture path. (B) Wet specimen surfaces were considerably rougher indicating an intergranular fracture path involving grain boundaries.
tom of the image to the top and as expected, the majority of the fatigue striations are aligned perpendicular to this. Ductile failure occurred when the fractured implant specimen was manually detached. Cup-like depressions, also known as dimples, are evident and are representative of this mode of fracture (Fig. 5C).
Fatigue striations leave a record of the incremental crack growth per cycle (da/dc). These steps became larger with distance away from the origin. Increment crack growth per cycle as a function of distance from the origin was curve-fit to choose a reasonable mathematical expression: y = a + b(ln x)
2
(1)
where y is the crack growth increment per cycle (m), x is distance from the failure origin (m) and a and b are fitted constants. This expression was chosen as being the simplest relationship that was well-fit to all data sets between 225 and 1400 m for both 2 and 30 Hz conditions (r2 between 0.52 and 0.77). Curve-fit graphs for all four conditions indicated that fatigue striations begin to appear approximately 225 m from the origin, consistent with the SEM observations. From approximately 750 to 1500 m from the origin, curve-fits
Table 1 – Calculation of the duration and percentage of cycles for fatigue failure based on incremental crack growth measurements on failed implants and Eq. (2) (D = dry, W = wet). Specimen 2.2D 2.4D 2.1W 2.3W 30.10D 30.3W a b
Total cycles to failure 229,364 288,619 214,672 79,926 2,297,637 3,918,266
Approx. cycles for fatigue crack growtha 1891 2851 4216 2326 1442 1442
Approx. time during fatigue crack growth (min) 15 24 35 19 1 1
Actual stepwise crack growth (i.e. fatigue) was found to be limited to these numbers of cycles. Lifetime represents the total number of cycles to failure.
Percentage of cycles for fatigue crack growth (% lifetime)b 0.8 1.0 2 2.9 0.1 0
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
1423
Fig. 6 – Incremental crack growth per cycle as a function of distance from the origin for a pair of 2 Hz specimens: (A) 2 Hz, wet; and, (B) 2 Hz, dry. Dots are measured data; inner line is the fitted curve to Eq. (1); outer lines are 95% confidence boundaries. Cracks are seen to be growing more slowly in the presence of saline.
fatigue crack growth:
Fig. 5 – Higher magnifications of fracture surfaces in Fig. 4: (A) crack pop-in stage of failure corresponding to location (A) in Fig. 4; (B) fatigue crack growth region corresponding to location (B) in Fig. 4; and, (C) ductile failure corresponding to location (C) in Fig. 4.
and 95% confidence intervals for incremental growth differ between 2 Hz wet (Fig. 6A) and 2 Hz dry (Fig. 6B). The distance between fatigue striations for 2 Hz wet become less than that for 2 Hz dry, suggesting a diminished incremental crack growth per cycle in the presence of the saline solution (i.e. slower crack growth). An estimate of the number of cycles needed for fatigue growth to the pre-defined failure point was determined by integrating the inverse of Eq. (1) between 225 and 1400 m to obtain an estimate of the total number of cycles representing
1400
225
1 y
(2)
It appears that fatigue crack growth occurred over approximately 1400–4200 cycles (Table 1). Compared to the total number of cycles to failure, greater than 97% of the lifetime of the implant was used to create an initial crack. The second method, of scanning the implant displacement per cycle data to estimate the number of cycles devoted to fatigue failure yielded similar results (Table 2). Fatigue crack growth occurred over approximately 1000–3600 cycles. This represents less than 1% of the lifetime of the implant specimens.
4.
Discussion
One observation from this work is that cycles to failure at one stress level are bi-modally distributed. This bi-modal distri-
1424
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
Table 2 – Calculation of the duration and percentage of cycles for fatigue failure based on load displacement data on failed implants (D = dry, W = wet). Specimen 2.2D 2.4D 2.1W 2.3W 30.10D 30.3W
Total cycles to failure 229,364 288,619 214,672 79,926 2,297,637 3,918,266
Approx. cycles for fatigue crack growth
Approx. time during fatigue crack growth (min)
1620 1140 1590 Data not available 3600 Not detectable
bution suggests that the implant population may contain two distinct groups, for example a certain percentage containing a major residual machining flaw and the remainder free of that specific defect. Alternatively the bi-modal distribution may be related to characteristics of the alloy and represent an outcome of more complex origins inherent to that particular implant under low-load cyclic testing. In either case larger numbers of implants likely need to be tested in order to more fully explore the full failure distributions for each subset. The need for sophisticated experimental designs and larger numbers of specimens may complicate future studies hoping to make statistical comparisons of clinical interest; e.g., comparing seemingly similar implants from two different manufacturers. Comparing the 2 Hz wet versus 2 Hz dry fracture surfaces, the wet specimens appear to have a smaller crack step per cycle. For wet specimens it appears that more crack growth was intergranular than transgranular and that secondary cracking events associated with grain boundaries were more frequent than for dry specimens. Perhaps the saline solution contributed to a chemically assisted corrosion process during fracture, allowing the applied energy to be dissipated into creating the lateral cracks in multiple planes; consequently, the advancement of the crack front was retarded. Another likely factor is a decreased stress intensity associated with cracks turning away from the fracture plane to follow grain boundaries. It appears from this work that initial crack pop-in was more likely to happen at 2 Hz than at 30 Hz. This finding is counter to previous work suggesting that higher stresses would develop in the root of threads as the “effective” elastic modulus of the base increases at 30 Hz versus 2 Hz (unpublished research from our laboratory dealt with in a companion paper [20]). That the base becomes stiffer with higher cycling rates is seen in a shift to smaller implant displacements, likely due to viscoelastic behavior of the filled epoxy base material. That this leads to higher stresses in the root of implant threads is a finding of finite element analysis. Another phenomenon may be competing with higher stresses under 30 Hz testing. One potential candidate would be “strain rate sensitivity”; under lower strain rates (2 Hz) there may be damage accumulation mechanisms that are not favored to occur at higher strain rates (30 Hz). Future work may involve direct measurement of strain rates near the implant surface at risk (as is the topic of a companion paper [20]). There has been minimal research on understanding implant fatigue, with studies mainly limited to case reports. Velásquez-Plata et al. [21] presented a clinically failed implant,
Percentage of cycles for fatigue crack growth (% lifetime)
14 10 13
0.7 0.4 0.7
2
0.2
claiming that SEM images revealed fatigue striations. However, there is a lack of peak and valleys that are characteristic of fatigue striations. The uniformly spaced lines that they describe appear more consistent with machining grooves. Piattelli et al. reported on three clinically fractured implants [22]. Their features described as fatigue striations appears more consistent with those obtained in our 2 Hz wet specimens. Morgan et al. examined clinically fractured implants and compared them to laboratory fractured implants [23]. Their images demonstrated fatigue striation spacing from 0.1 to 1.0 m consistent with our measurements. They did not confirm our observation that the fatigue crack front changed directions on a local scale likely involving crystallographic grain orientations. All of our observations involving grain orientations and grain boundaries require further validation via analysis of the microstructure of the implants tested.
5.
Conclusions
Preliminary analysis suggests that testing in air and normal saline are equivalent in terms of likelihood of fracture versus runout. Implant fatigue failure involved three distinct stages (all of which can be visualized by SEM along with some quantitative measures): (1) brittle crack pop-in and arrest, (2) fatigue crack growth, and (3) final ductile failure. Failures were found to be bi-modally distributed, either (1) <350,000 cycles or (2) >1.5 million cycles, likely limiting future comparative analysis to subgroups representing different failure mechanisms. On a microscopic level, fatigue crack growth rates appeared to be similar under 2 and 30 Hz testing, but may be different for wet and dry conditions at 2 Hz. The failure process occurred over a very limited number of cycles; the majority of the cycles were used to pop-in the initial crack with catastrophic failure then occurring after just a few thousand cycles of fatigue crack growth. Much work has gone into studying the fracture surfaces of failed implants by SEM. Continuations of this effort may lead to the development of powerful tools for understanding whether different test protocols and different laboratories are producing failure by similar mechanisms. Eventually such tools may be useful in examining clinically failed specimens for the validation of laboratory tests.
Acknowledgements This project was supported by a grant from the ITI Foundation for the Promotion of Oral Implantology, Switzerland.
d e n t a l m a t e r i a l s 2 5 ( 2 0 0 9 ) 1419–1425
references
[1] Taylor TD. Prosthodontic problems and limitations associated with osseointegration. J Prosthet Dent 1998;79:74–8. [2] Zarb GA, Schmitt A. The longitudinal clinical effectiveness of osseointegrated dental implants: the Toronto study. Part III. Problems and complications encountered. J Prosthet Dent 1990;64:185–94. [3] Tolman DE, Laney WR. Tissue-integrated dental prosthesis complications. Int J Oral Maxillofac Implants 1992;7:477– 84. [4] Naert I, Quirynen M, van Steenberghe D, Darius P. A study of 589 consecutive implants supporting complete fixed prostheses. Part II. Prosthetic aspects. J Prosthet Dent 1992;68:949–56. [5] Pylant T, Triplett RG, Key MC, Brunsvold MA. A retrospective evaluation of endossesous titanium implants in the partially edentulous patient. Int J Oral Maxillofac Implants 1992;7:195–202. [6] Gunne J, Jemt T, Linden B. Implant treatment in partially edentulous patients: a report on prostheses after 3 years. Int J Prosthodont 1994;7:143–8. [7] Lekholm U, Gunne J, Henry P, Higuchi K, Lindén U, Bergström C, van Steenberghe D. Survival of the Branemark implant in partially edentulous jaws: a 10-year prospective multicenter study. Int J Oral Maxillofac Implants 1999;14:639–45. [8] Henry PJ, Laney WR, Jemt T, Harris D, Krogh PH, Polizzi G, Zarb GA, Herrmann I. Osseointegrated implants for single tooth replacement: a prospectve 5-year multicenter study. Int J Oral Maxillofac Implants 1996;11:450–5. [9] Balshi TJ. An analysis and management of fractured implants: a clinical report. Int J Oral Maxillofac Implants 1996;11:660–6. [10] Allen PF, McMillan AS, Smith DG. Complications and maintenance requirements of implant-supported prostheses provided in a UK dental hospital. Br Dent J 1997;182:298–302. [11] Adell R, Lekholm U, Rockler B, Brånemark PI. A 15-year study of osseointegrated implants in the treatment of the edentulous jaw. Int J Oral Surg 1981;10:387–416.
1425
[12] Goodacre CJ, Kan JY, Rungcharassaeng K. Clinical complications of osseointegrated implants. J Prosthet Dent 1999;81:537–52. [13] Eckert SE, Meraw SJ, Cal E, Ow RK. Analysis of incidence and associated fractures with fractured implants; a retrospective study. Int J Oral Maxillofac Implants 2000;15:662–7. [14] Norton MR. An in vitro evaluation of the strength of an internal conical interface compared to a butt joint interface in implants design. Clin Oral Implants Res 1997;8:290–8. [15] Möllersten L, Lockowandt P, Linden L. Comparison of strength and failure mode of seven implant systems: an in vitro test. J Prosthet Dent 1997;78:582–91. [16] Khraisat A, Stegaroiu R, Nomura S, Miyakawa O. Fatigue resistance of two implant/abutment joint designs. J Prosthet Dent 2002;88:604–10. [17] Perriard J, Wiskott WA, Mellal A, Scherrer SS, Botsis J, Belser UC. Fatigue resistance of ITI implant-abutment connectors—a comparison of the standard cone with a novel internally keyed design. Clin Oral Implants Res 2002;13:542–9. ˘ H, S¸ahin S. Dynamic [18] C¸ehreli MC, Akc¸a K, I˙plikc¸ioglu fatigue resistance of implant-abutment junction in an internally notched morse-taper oral implant: influence of abutment design. Clin Oral Implants Res 2004;15:459–65. [19] Wiskott HW, Nicholls JI, Belser UC. Stress fatigue: basic principles and prosthodontic implications. Int J Prosthodont 1995;8:105–16. [20] Karl M, Kelly JR. Influence of loading frequency on implant failure under cyclic fatigue conditions. Dent Mater 2009;25:1426–32. [21] Velásquez-Plata D, Lutonsky J, Oshida Y, Jones R. A close-up look at an implant fracture: a case report. Int J Periodontics Restorative Dent 2002;22:482–91. [22] Piatelli A, Piatelli M, Scarano A, Montesani L. Light and scanning electron microscopic report of four fractured implants. Int J Oral Maxillofac Implants 1998;13:561–4. [23] Morgan MJ, James DF, Pilliar RM. Fracture of the fixture component of an osseointegrated implant. Int J Oral Maxillofac Implants 1993;8:409–14.