See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226416080

Acceptance Criteria for Corrosion Resistance of Medical Devices: Statistical

Analysis of Nitinol Pitting in In Vivo Environments

Article  in  Journal of Materials Engineering and Performance · August 2009

DOI: 10.1007/s11665-009-9420-z

CITATIONS READS

16 1,732

7 authors, including:

Larry Eiselstein Andrew Nissan

Exponent Tesla Motors
67 PUBLICATIONS   308 CITATIONS    9 PUBLICATIONS   40 CITATIONS   

SEE PROFILE SEE PROFILE

Roberto Dugnani
Shanghai Jiao Tong University
54 PUBLICATIONS   313 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

atmospheric marine corrosion View project

Relating fractographic features in brittle materials to aspects of the dynamic crack propagation View project

All content following this page was uploaded by Larry Eiselstein on 29 October 2014.

The user has requested enhancement of the downloaded file.

JMEPEG ÓASM International
DOI: 10.1007/s11665-009-9420-z 1059-9495/$19.00

Acceptance Criteria for Corrosion Resistance

of Medical Devices: Statistical Analysis
of Nitinol Pitting in In Vivo Environments
Lawrence E. Eiselstein, Duane Steffy, Andrew Nissan, Nigel Corlett, Roberto Dugnani, Esra Kus, and Sarah G. Stewart

(Submitted October 24, 2008; in revised form February 23, 2009)

ASTM F 2129 test method nor the FDA provides any guidance as to what constitutes an acceptance
criterion for the corrosion resistance of implantable medical devices. Neither provide any guidance on how
many samples to test or how to handle censored data, i.e. datasets where there are only a few tests that
breakdown. The development of both a statistically valid acceptance criterion for corrosion resistance and a
method of evaluation would be of significant benefit to the medical device community. This study of 420
nitinol cyclic polarization tests, which builds on previous research that was presented at SMST 2007,
investigates the effect of long-term exposure to simulated in vivo environments with differing degrees of
aeration. This was accomplished by pre-exposing electropolished (EP) nitinol to phosphate buffered saline
(PBS) at 37 °C that had been sparged with either ultra high purity nitrogen or laboratory air. Immersion
times ranged from 1 h up to 30 days. A total of 290 EP samples were tested in order to obtain a reasonable
number of samples with breakdown, i.e. pitted. In addition, a total of 130 mechanical polished (MP)
samples were also analyzed. This data allow us to test our statistical model that was presented at SMST
2007. This model takes into account the probability of breakdown per unit of exposed surface area and, if
breakdown occurs, predicts the probability that Eb 2 Er is greater than some threshold value. Aerated PBS
environments were found to have a large influence on the margin of safety against pitting in vivo. Statistical
methods for treating highly right censored pitting data are presented.

However, neither the ASTM F 2129 test method nor the

Keywords ASTM F 2129, biomaterials, corrosion testing, elec-
trochemistry, nitinol, pitting, quality assurance FDA provides any guidance as to what constitutes an
acceptance criterion for the polarization results or how many
samples should be tested. ASTM F 2129 suggests comparison
with a device already approved and on the market in the
United States that is without any postmarket approval
1. Introduction incidents of corrosion failures. This approach can, however,
be quite expensive as it requires testing twice the number of
experimental units (i.e., number of candidate devices and
Long-term reliability of implantable medical devices requires
predicate devices); moreover, it presumes the existence of a
testing to assure low corrosion and metal ion release rates.
suitable predicate device, i.e. a device that is implanted in a
Corrosion can reduce the structural integrity of the device and the
similar location in the body and manufactured from similar
release of metal ions has been associated with adverse immu-
material and that the medical-device company has accurate
nological response (Ref 1-3). ASTM F 2129 is a test method used
knowledge regarding the predicate device
s in vivo corrosion
to assess the corrosion susceptibility of small metallic implants
resistance.
using cyclic potentiodynamic polarization (Ref 4). The United
As discussed in our earlier paper on nitinol corrosion, NiTi
States Food and Drug Administration (FDA) typically requires
appears to have very good corrosion resistance based on many
medical-device manufacturers to provide data that will demon-
patient-years of in vivo experience (Ref 5). Nevertheless, the
strate that their device will have sufficient corrosion resistance
development of both a statistically valid acceptance criterion
during the expected in vivo service life which could be greater
for corrosion resistance and a method of evaluation would be
than 10 years. Frequently, medical-device manufacturers use
of significant benefit to the medical-device community. In
ASTM F 2129 to satisfy this FDA requirement.
addition, there is a need for a protocol by which a medical-
device manufacturer can periodically check the corrosion
This article is an invited paper selected from presentations at Shape
resistance of an approved medical-device to assure that the
Memory and Superelastic Technologies 2008, held September 21-25, manufacturing process is still within specification and devices
2008, in Stresa, Italy, and has been expanded from the original are being manufactured with adequate corrosion resistance, i.e.
presentation. a need for a manufacturing lot acceptance criterion. This study
Lawrence E. Eiselstein, Duane Steffy, Andrew Nissan, Nigel Corlett, is an attempt to establish such a methodology and is a
Roberto Dugnani, Esra Kus, and Sarah G. Stewart, Exponent, Menlo continuation of work that was first reported at SMST 2007 in
Park, CA. Contact e-mail: [email protected]. Japan (Ref 5).

Journal of Materials Engineering and Performance

 1.000

0.800

0.600

Potential in V (SCE)
0.400

Eb
0.200

0.000

-0.200
E b - Er

-0.400 Er
-0.600
1E-11 1E-10 1E-09 1E-08 1E-07 1E-06 1E-05 1E-04 1E-03 1E-02
Current in A

Fig. 1 Definition of electrochemical parameters Eb, Er, and Eb
Er

Three general methodologies exist as possibilities for an 1.8

acceptance criterion. The first is that the pitting or breakdown Er
1.6 Eb
potential (Eb), a measure of resistance to localized corrosion in
a specific environment, of an implantable device should have 1.4
similar or better corrosion resistance as determined by ASTM F
Probability density

1.2
2129 testing than approved devices currently on the market
with no known corrosion problems. The electrochemical 1.0
µ(Eb-Er)
parameters discussed in this paper are shown in Fig. 1. These 0.8
exemplar devices should ideally be manufactured from similar
materials and be used in the same general location in the body 0.6
as proposed for the current device to assure a similar in vivo
0.4
exposure environment, i.e. exposure to blood, bile, urine,
saliva, etc. As mentioned above, it is sometimes difficult, and 0.2
expensive, to find such exemplars and there is little guidance on
0.0
how many devices should be tested to confirm equivalent or -0.2 0.0 0.2 0.4 0.6
better corrosion resistance. The second methodology that has Electrochemical Potential (V/SCE)
been proposed is that the breakdown potential (Eb) of the
device should be greater than some threshold value, indepen- Fig. 2 Schematic illustrating the margin of safety against pitting
dent of the material used for the implant (Ref 6-8). These
implant alloys can have drastically different pitting resistance
and it is somewhat arbitrary that a minimum Eb should be nearly zero or negative then the device will most likely pit
specified independent of the material and environment. It is when exposed to that environment. This is illustrated in Fig. 2.
generally agreed that if a NiTi medical-device can be made This Eb
Er parameter has been used frequently in the past to
reliably with an Eb of greater than 0.600 V with respect to a minimize the risk of pitting in many industries (Ref 9-13).
saturated calomel electrode (SCE), then there is little doubt that One problem with the Eb
Er criteria, as discussed by
these materials will be resistant to corrosion in vivo unless Rosenbloom and Corbett, is that there is little information
other conditions like fretting, galvanic effects, or crevice regarding Er for in vivo conditions (Ref 8). Another issue is
corrosion take place in vivo. The reality, however, is that many how Eb, Er, and ultimately Eb
Er vary during long-term in
medical devices will not be able to achieve such high values vivo exposures. The current ASTM F 2129 Standard states that
due to material selection, material finish, design, and manu- devices are to be tested in their finished state. However, it is
facturing limitations. The third methodology, which we further unlikely that the device, after exposure to in vivo conditions for
develop in this paper, is to use the difference between the several months will have the same resistance to pitting
breakdown potential (Eb) and the open-circuit or rest potential corrosion or even the same rest potential as a result of the
(Er). This gap (Eb
Er) is considered to be a better parameter to surface film modification. Finally, there seems to be a wide
evaluate medical-device margins against pitting in vivo as has variation in Eb values for NiTi depending on surface finish.
previously been discussed by Pound (Ref 9). The basic Some NiTi:
principle behind this approach is that if the in vivo Er is much
lower than the pitting potential the device will have a large 1. will always pit below the minimum potential of 800 mV
margin of safety against pitting or correspondingly, if Eb
Er is specified in ASTM F 2129;

Journal of Materials Engineering and Performance

2. will never undergo pitting below 800 mV; and Auger analysis revealed surface oxide thicknesses of
3. will infrequently breakdown. 28.6 ± 3.7 and 5.72 ± 0.79 nm for MP and EP wire, respec-
tively (Ref 15). These film thicknesses were based on four
Normal quality-control methodologies can be used for measurements from two different locations on two different MP
samples that always pit or never pit, but case 3, samples that and EP NiTi wires. The thickness measurement was based on
only infrequently breakdown (datasets that contain censored the equivalent sputter rate of SiO2 under identical conditions
data), are more difficult to treat statistically. used for the NiTi. Oxide thickness measurements were
In an attempt to answer some of these questions, and to calculated as the thickness where the oxygen intensity dropped
investigate whether long-term in vivo exposure might affect the to half of its maximum within the profile. Both oxide films on
margin of safety against localized corrosion, we report how the MP and EP had a 2 wt.% or less Ni concentration at the
long-term immersion in aerated and nonaerated phosphate surface (Ref 15). All electrochemical testing was done with
buffered saline (PBS) will affect Eb, Er, and ultimately Eb
Er 2.5-cm long wire. Both cut ends of the wire were coated with
on a statistically significant number of samples with two silicone sealant which resulted in a surface area of 1.15 cm2.
different surface conditions: mechanically polished (MP) and Figure 3 shows a representation of the size distribution of
electropolished (EP). inclusions from longitudinal and transverse cross sections of
Experimental data were collected on the breakdown the wire, and is similar for both materials used. The longitu-
potential (Eb) of replicate surrogate devices (i.e., wire samples) dinal sections indicate that the inclusions are present as
and on the variation in rest potential (Er) with immersion time stringers that broke up during the drawing process and are
for individual units. Using these data, we conducted simulation relatively strong and brittle compared to the surrounding
studies to investigate the implications for device perfor- matrix, as indicated by the fracturing and voids formed around
mance—specifically, the probability that the rest potential these inclusions. We were unable to classify the inclusions with
would exceed breakdown potential during the life of the device. energy dispersive spectroscopy (EDS) but were able to observe
The simulation results provided statistical support in defining a that they are titanium rich. Infrequently, these inclusions
threshold value that would represent an acceptable margin intersect the wire surface as is shown in Fig. 4 for an EP
between breakdown and rest potential (Eb
Er), as well as a specimen. Locating surface inclusions on the MP wire was
correspondingly low probability of corrosion-induced failure much more difficult due to the rough surface and thick oxide
during the lifetime of the device. layer.
We then implemented a tolerance interval approach for All pre-test immersions and electrochemical testing were
determining, from standard testing of a candidate device, performed in PBS (8.0 g L
1 NaCl, 0.2 g L
1 KCl, 1.15 g L
1
whether there is sufficient confidence that a high proportion of Na2H2PO4, and 0.2 g L
1 KH2PO4) at 37 ± 1 °C. The pH of
replicate units will attain the specified threshold value. Finally, the solutions was 7.4 ± 0.1 for all immersions and electro-
we developed guidance for medical-device manufacturers
chemical testing. When samples were pre-exposed to aerated
quality-assurance programs in addressing the key practical PBS, these solutions were aerated with laboratory air (79% N2,
question of how to determine the number of replicate units 21% O2, and 0.038% CO2). For deaerated tests, the PBS was
required for testing in order to establish an acceptable lower sparged with ultrahigh-purity nitrogen for 30 mins prior to
bound on the value of Eb
Er for manufactured devices. testing and then continuously throughout the test. The use of
Developing this lower bound can be quite challenging for NiTi deaerated and aerated solutions was chosen to represent the
materials with good surface treatments since the vast majority widest range of oxidizing conditions to which an implantable
of samples do not exhibit breakdown. We suggest a method- medical-device might be subjected. Obviously, the deaerated
ology to treat this type of highly censored data, i.e. few PBS has a lower redox potential than would be expected to be
breakdowns with a large number of censored data, i.e. values found in vivo, and the aerated PBS is expected to provide
greater than 800 mV SCE. higher redox potentials than conditions in vivo, except for
perhaps implants in the airways or devices that are subjected to
peroxides from an immunological response.
All tests were performed using Gamry computer-controlled
2. Materials and Methods potentiostats and Gamry DC-105 software. Long-term potential
monitoring was performed in accordance with ISO 16429:2004
This investigation evaluated the pitting resistance of NiTi and the cyclic polarization tests were performed in accordance
wire in two as-received surface conditions: MP and EP. The with ASTM F 2129-06 (Ref 4, 16). All potentials were
surface preparation of the material is proprietary to the material measured with respect to a SCE. The potentiodynamic scan rate
supplier and was not provided. The EP and MP NiTi used in the used was 0.167 mV/s and the potential increased until break-
experiments came from different material lots. down occurred or 1.1 V (SCE) was reached. All Er values
The NiTi properties and surface characterization for the MP reported are the corrosion potentials observed at the end of the
specimens that were purchased were described 55.9%Ni pre-test immersions, i.e. in aerated or deaerated solutions, not
content superelastic wire with an Af temperature in the the Er measured in the deaerated PBS used to determine Eb.
5-15 °C temperature range with a bright finish described as
mechanical polishing that was followed by an argon anneal.
The EP specimens that were purchased were described as
superelastic straight wire that conforms to ASTM F 2063 3. Results
Specifications (Ref 14). The EP materials had a nominal
composition of 55.8 wt.% Ni. This material had an Af We tested a total of 420 nitinol test specimens in order to get
temperature in the 5-18 °C range. Both the MP and EP wire sufficient information regarding the probability of pitting in 130
had a diameter of 1.45 mm (0.057 in). mechanical polished and 290 EP NiTi samples. The pre-test

Journal of Materials Engineering and Performance

Fig. 3 Inclusions from metallurgical cross sections. Longitudinal (upper left) and transverse (upper right) of EP wire. The corresponding longi-
tudinal and transverse metallographic sections from the MP wire are shown at the bottom left and right

Fig. 4 Inclusions intersecting the surface. EP (left) and MP (right) in as-received condition

exposure times to the deaerated and aerated environments were is about the same and unaffected by oxygen concentration in
1, 12, 60, 156, and 720 h for the MP surface finish and 1, 12, the PBS as shown in Fig. 6. However, Eb does increase with
and 60 h for the EP samples. Then, 3-99 replicate specimens increasing pre-test exposure time. The safety margin against
were tested at each pre-test immersion time. The results from pitting corrosion, Eb
Er, is lower in aerated environments
our testing are summarized in Fig. 5-7 and Table 1. The mean primarily from the effect of oxygen on raising Er (Fig. 7).
and standard deviation for Eb for EP samples can only be
calculated for the samples that pitted and therefore underesti-
mate the true resistance of the material to pitting. As can be
seen from Table 1, there are very clear differences between the 4. Discussion
corrosion resistance of the MP and EP samples. All of the MP
samples exhibited pitting below 0.8 V (SCE) whereas only All the MP samples pitted, whereas only about 13% of the
13% of the EP samples pitted. Figure 5 illustrates that EP samples pitted indicating a significant effect of surface
increasing oxygen concentration in the PBS increases Er and treatment. The dissolved oxygen concentration also has a very
that Er also tends to increase with increasing pre-test exposure large influence, as expected on the open-circuit potential (Er).
time in both deaerated and aerated PBS solutions. Er increases The MP wire had an oxide film about five times as thick as the
less rapidly with increasing pre-test exposure time for the EP EP samples and had significantly lower corrosion resistance as
surface condition. Eb for either the MP or EP surface condition seen in Table 1. Clarke et al. showed that NiTi with a thicker

Journal of Materials Engineering and Performance

 0.2 0.2

Rest Potential, Er (V/SCE)

Rest Potential Er (V/SCE)

0.0 MP aerated 0.0
EP Aerated

-0.2 -0.2

-0.4 MP deaerated
-0.4 EP Deaerated

-0.6 -0.6

-0.8 -0.8
0.1 1 10 100 1000 0.1 1 10 100 1000
Pre-exposure time (hours) Pre-exposure time (hours)

Fig. 5 Effect of pre-test immersion on rest potential Er

1.2 1.2
Eb MP in deaerated PBS at 37°C
1.0 Eb MP in aerated PBS at 37°C 1.0 EP deaerated
EP aerated

0.8 0.8
Eb (V/SCE)

Eb (V/SCE)
0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
10
0
10
1
10
2
10
3 0.1 1 10 100 1000
Pre-exposure time (hours) Pre-exposure time (hours)

Fig. 6 Effect of pre-test immersion on break down potential Eb

Table 1 Summary of test results

Er Eb Eb 2 Er
Time, No. No. % No. incomplete
NiTi Gas h RPT (a) Mean r No. CPT (b) pitted Pitted Mean r Mean r tests (c)

MP N2 1 31
0.472 0.062 30 30 100 0.130 0.047 0.602 0.078 1

MP Air 1 13
0.160 0.035 13 13 100 0.138 0.067 0.298 0.076 0
MP N2 12 27
0.476 0.110 27 27 100 0.095 0.081 0.571 0.095 0
MP Air 12 6
0.104 0.009 6 6 100 0.246 0.065 0.350 0.069 0
MP N2 60 16
0.307 0.161 16 16 100 0.212 0.158 0.520 0.153 0
MP Air 60 13
0.015 0.049 11 11 100 0.283 0.018 0.310 0.038 2
MP N2 156 15
0.090 0.082 15 15 100 0.393 0.098 0.482 0.168 0
MP Air 156 6 0.024 0.011 6 6 100 0.306 0.059 0.281 0.062 0
MP N2 720 3
0.049 0.113 3 3 100 0.530 0.131 0.578 0.129 0
MP Air 720 0 ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ
EP N2 1 98
0.265 0.082 97 17 18 0.585 0.195 0.832 0.201 1
EP Air 1 58
0.180 0.069 57 10 18 0.407 0.184 0.584 0.216 1
EP N2 12 58
0.245 0.060 56 4 7 0.590 0.193 0.769 0.198 2
EP Air 12 57
0.159 0.059 55 7 13 0.609 0.259 0.763 0.267 2
EP N2 60 13
0.192 0.093 12 0 0 ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ 1
EP Air 60 6
0.099 0.045 6 1 17 1.090 ÆÆÆ 1.199 ÆÆÆ 0
EP N2 156 0 ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ
EP Air 156 0 ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ ÆÆÆ

(a) No. RPT: Number of rest potential tests; not all samples had valid cyclic polarization tests
(b) No. CPT: Number of cyclic polarization tests; not all samples had valid cyclic polarization tests
(c) Not all samples had valid cyclic polarization tests

Journal of Materials Engineering and Performance

oxide layer has a lower Eb value and a greater nickel release in Shibata has shown that a normal distribution adequately
both long-term and short-term experiments (Ref 1). Clarke et al. describes the distribution of Eb for AISI 304 stainless steel
also found that after long-term immersion tests breakdown tested in 3.5 wt.% NaCl solutions at 30 °C (Ref 17, 18).
potentials increased for samples that exhibited lower values Nevertheless, Kowaka notes that pitting potentials are actually
initially (Ref 1). the lowest breakdown potential that is found only at a small
The effect of long-term pre-exposures of MP NiTi to 37 °C area on a sample under test, and as such, a ‘‘weakest link’’
PBS clearly increases Eb; however, Eb increases faster for pre- statistical distribution such as a Gumbel or Weibull distribution
test exposures to aerated solutions compared to deaerated is generally expected to fit such data (Ref 18). Kowaka showed
solutions. This can be seen by comparing Eb values for various that the Weibull distribution also provided a good fit to
pre-test exposure immersion times for MP NiTi in Table 1 and Shibata
s stainless steel Eb data while at the same time being a
Fig. 6. Clarke et al. found a similar trend in Eb for NiTi with a ‘‘weakest link’’ type of statistical distribution. The Eb data from
thick oxide (340 nm) exposed to 0.9 wt.% saline for up to the MP and EP specimens exposed to deaerated PBS are shown
4 months although they do not report how Er values were in Fig. 8. Both datasets are well described by a three-parameter
affected (Ref 1). Our data show that both Er and Eb increased Weibull distribution.
with increasing pre-immersion duration in deaerated PBS
(Fig. 5 and 6), but the trends were not equitable as Eb
Er
decreased with increasing immersion duration because Er
increased faster than Eb. 5. MP Modeling
The pre-test exposure of MP samples to aerated PBS had a
more dramatic effect than the longest exposure to deaerated Regression analysis of Eb
Er for MP NiTi data was
PBS with the Eb
Er values being about 300 mV lower which performed to quantify the effects of aeration and immersion
is primarily due to the very large increase of Er when exposed time. The effect of pre-test immersion exposure time was
to aerated PBS. modeled as linear in the logarithm scale and separate
The statistical distribution of Eb for stainless steels has regression analyses were performed on the aerated and
previously been studied by Shibata and others (Ref 17-19). deaerated data because of the greater variability of the

1.2 1.2
Eb-Er for MP in deaerated PBS at 37°C
1.0 Eb-Er for MP in aerated PBS at 37°C 1.0
Eb-Er (volts vs SCE)

Eb-Er (V/SCE)

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Eb-Er for EP in deaerated PBS

Eb-Er for EP in aerated PBS

0.0 0.0
1 10 100 1000 0.1 1 10 100 1000
Pre-exposure time (hours) Pre-exposure time (hours)

Fig. 7 Effect of pre-test immersion on Eb
Er

Fig. 8 Three-parameter Weibull distribution of Eb for MP (left) and EP (right) in deaerated PBS for 1 h

Journal of Materials Engineering and Performance

Table 2 Regression analysis results of Eb 2 Er oxygen concentration than a deaerated solution. The value
for MP NiTi provided above of
209 mV for the expected decrease in Eb
Er
after 10-years exposure to deaerated solutions is an extreme
SE extrapolation of the data and therefore may not accurately predict
Predictor Coefficient coefficient T P
actual in vivo performance. However, if testing is done as we
suggest, i.e. pre-exposure to aerated solutions, the inherent
Deaerated
Constant 0.6037 0.01962 30.77 <0.001 variability in Eb
Er is decreased and there is no need to account
ln, t
0.01838 0.006081
3.02 0.003 for the time the implant will be in service.
S = 0.1179 R2 = 9.3% R2 (adjusted) = 8.3% The probability that an MP sample will have a negative
Aerated Eb
Er can be calculated using the probability distribution
Constant 0.3083 0.0172 17.92 <0.001 statistics shown in Table 3 where we have combined both the 1
ln, t
0.00037 0.00533
0.07 0.945 and 12 h* and 60 and 156 h** deaerated data since there was no
S = 0.0649 R2 = 0.0% R2 (adjusted) = 0.0% statistically significant difference between the 1 and 12 h or the
60 and 156 h data; however, the difference between the 1 and
deaerated data. The fitted regression equation for the deaerated 12 h and the 60 and 156 h data was statistically significant. We
data is as follows: also calculated the probability that Eb
Er will be negative for
MP NiTi for the combined 60 and 156 h immersions in deaerated
Eb
Er ðV=SCEÞ ¼ 0:604
0:0184 ln½t PBS and the probability of negative Eb
Er for MP NiTi exposed
Ninety-one data points were used in this regression and the to aerated PBS for all our aerated data (1-156 h data combined)
ANOVA results are reported in Table 2. since Eb
Er versus pre-test exposure time is relative insensitive
Similarly for the aerated data the fitted regression equation is to exposure duration in aerated environments.
as follows: Table 3 shows the results using several different types of
distributions. A normal distribution is used since it has been
Eb
Er ðV=SCEÞ ¼ 0:308
0:0053 ln½t previously used to describe Eb data for AISI 304 stainless steel
Thirty-six data points were used in this regression and the in 3.5 wt.% saline (Ref 17-19). A Gumbel extreme value
ANOVA results are reported in Table 2. distribution has also been used since it theoretically should
Pre-test exposure to aerated environments has a large effect on model the distribution of smallest values for multiple measure-
the corrosion margin of safety primarily due to the large effect of ments taken in different locations of constant area (Ref 18).
aeration on Er as can be seen by the constant term dropping from Among a set of candidate models that included the two- and
604 to 308 mV. Table 2 shows that the ln(t) factor is statistically three-parameter Weibull and lognormal distributions, the
significant for deaerated samples but not for aerated samples. Gumbel extreme value distribution was found to be the best
However, the low R2 value for the deaerated samples indicates fit to the Eb
Er data for MP samples pre-exposed to deaerated
the presence of large unexplained variability. We suspect that this PBS for 1 and 12 h (Fig. 9). The superiority of the Gumbel
large variation is a result of the inherent randomness of the pitting versus the normal model is reflected in the larger likelihood
process. The average Eb
Er value for the longest duration pre- value of 65.9 in Table 3. This finding is also reflected in the
test immersion (720 h) appears to violate this decreasing trend, more reasonable (higher) pitting probability predicted using the
but the average is calculated from only three observations and Gumbel distribution (6.43 9 10
5 versus 5.14 9 10
12). A
does not alter the statistical significance of the trend. three-parameter Weibull distribution was found to best charac-
For deaerated samples, the baseline for Eb
Er is defined as terize the longer-term deaerated immersion data (60-156 h)
material subjected to pre-test exposure for 1 h. The regression with an estimated pitting probability of zero versus a
equation for deaerated samples can therefore be expressed as: 6.66 9 10
4 probability if a normal distribution is assumed.
The three-parameter Weibull distribution was also found to
Eb
Er ðV=SCEÞ best characterize the aerated Eb
Er data. This is not surprising
¼ baseline þ pre-test immersion duration effect since the two- and three-parameter Weibull distributions are
typically among the probability models providing the best fit to
where baseline = 604 mV (SCE); this is the Eb
Er from the minimum extreme value distributions such as Eb. The three-
model for NiTi in the as-received condition (i.e., pre-test parameter Weibull model gives an estimated probability of
exposure to deaerated PBS for 1 h). pitting in vivo of zero whenever the threshold parameter—i.e.
Pre-test immersion duration term =
0.0184ln[t]; the model the minimum observable value of Eb
Er—is estimated to be
predicts a decrease in Eb
Er as the pre-exposure immersion positive. For example, the Eb
Er threshold is estimated at
time increases: 149 mV (SCE) for the samples exposed to aerated solution and
at 224 mV (SCE) for the deaerated exposures. Although both
12 h =
30 mV distributions would predict zero probability of pitting in vivo,
60 h =
96 mV the threshold values suggest that the specimens exposed to
156 h =
119 mV aerated PBS have a lower margin of safety against pitting. The
720 h =
121 mV fit of three-parameter Weibull distributions to Eb
Er data for
10-years =
209 mV short-term and long-term exposure to deaerated PBS are shown
in Fig. 9. Similarly, Fig. 10 shows the three-parameter Weibull
Sparging with air provides dissolved oxygen levels that are fit to the aerated Eb
Er data (all exposure times included).
higher than those generally found in vivo and therefore the
aeration effect here is exaggerated over that which would be *Mann-Whitney Rank Sum Test (P = 0.101).
expected when NiTi is exposed to arterial blood. Nevertheless, an **Mann-Whitney Rank Sum Test (P = 0.385).

aerated PBS solution better represents the in vivo dissolved Mann-Whitney Rank Sum Test (P = 0.001).

Journal of Materials Engineering and Performance

Table 3 Estimated probabilities that Eb 2 Er < 0 for MP NiTi
Solution Distribution type Immersion time, h Point estimate (PE) 95% Upper bound (PU) Log-likelihood

Deaerated Normal 1-12 5.14 9 10
12 4.96 9 10
9 58.8

Deaerated Gumbel (extreme value) 1-12 6.43 9 10
5 3.60 9 10
4 65.9
Deaerated Normal 60-156 6.66 9 10
4 6.66 9 10
3 13.6
Deaerated 3-parameter Weibull 60-156 0 (for Eb
Er < 224 mV) 0 16.0
Aerated Normal 1-156 5.57 9 10
7 5.07 9 10
5 48.4
Aerated 3-parameter Weibull 1-156 0 (for Eb
Er < 149 mV) 0 49.5

Fig. 9 Distribution of Ep
Er with confidence bands for MP samples immersed in deaerated PBS for 1 and 12 h as Gumbel (left) and
immersed in deaerated PBS for 60-156 h as three-parameter Weibull (right)

The entries in Table 3 clearly demonstrate that estimated

failure probabilities depend very sensitively on the type of
probability distribution used to model the observed variation in
Eb
Er. When a typical amount of testing is performed, several
models may fit the available data and lead to greatly differing
inferences regarding extreme values. The normal distribution
fits the observed test data reasonably well but, in all cases
studied, is not the best distribution to describe the distribution
of Eb
Er data. Specifically, Table 3 shows that, if the normal
model is used to calculate the probability of pitting in vivo, it
leads the counter intuitive conclusion that pitting is less likely
in aerated versus deaerated solutions. This result can be
explained by the short tails of the normal distribution and the
fact that, although Eb
Er, on average, is greater under
deaerated conditions, the associated variability about the mean
is also much greater than that under aerated conditions. The
Weibull model provides a marginally better fit to the data but
effectively precludes the possibility that Er could exceed Eb.
Use of the Gumbel extreme value distribution can be justified
by physical considerations and produces conservative estimates
of lower-tail probabilities, but the fit of the model to observed
data is poor in two of the three cases studied.

6. EP Modeling

A total of 290 EP samples were tested (169 in deaerated Fig. 10 Distribution of Ep
Er with confidence bands for MP sam-
PBS and 121 in aerated PBS). Analysis of variance of the EP ples immersed in aerated PBS for 1-156 h (three-parameter Weibull)

Journal of Materials Engineering and Performance

rest potential data (see Table 4) indicates that Er is significantly samples did not break down. One way to treat such data is
affected by both the amount of dissolved oxygen in the PBS through the use of nonparametric statistical methods. However
and the length of time of immersion. The effect of dissolved such methods require a large number of samples. For instance,
oxygen on Er is the combined effect of the dissolved on to specify a 99% reliability value with 95% confidence, 299
solution redox potential and the effect of long-term exposure on samples need to be tested. The number of specimens required
the passive film (morphology, composition, and thickness). for a nonparametric approach is much greater than a parametric
Of the 290 EP samples tested, only 39 (13%) pitted below approach (Ref 20).
0.800 V (SCE) the minimum voltage specified in ASTM F One parametric approach is to fit the censored Eb
Er
2129 at which the cyclic polarization test can be stopped. values to a standard statistical distribution using a methodology
Table 5 shows how probability of pitting of EP samples varies that can handle the censoring. We have done this for the aerated
with pre-test exposure time and oxygen concentration in the and deaerated EP specimens as shown in Fig. 11. These data
PBS. The results from Fisher
s exact test (Table 5) and binary were best fit by a three-parameter Weibull distribution for both
logistic regression analysis using the censored data versus the the aerated and deaerated exposures. The threshold parameter
oxygen and time variable (Table 6) showed that the differences (gamma), below which the probability of pitting is zero, is
observed were not significant indicating that neither oxygen estimated at 512 mV for the deaerated and 286 mV for the
concentration or duration of immersion in PBS significantly aerated exposures. These results suggest that aerated environ-
affect the percentage of EP samples that pit below 1.1 V (SCE). ments are more prone to produce pitting than deaerated
Modeling of the EP test data is more challenging than for the solutions, but neither solution is likely to result in breakdown
MP test data as most of the data are censored, i.e. most of the in vivo since the estimated probability of failure is zero for both.
The prediction of a zero failure probability from a fitted three-
Table 4 Analysis of variance of Er data from EP samples parameter Weibull or LogNormal distribution, for instance, does
not seem physically reasonable, since there is presumably some
General linear model: Er vs. oxygen concentration and time probability of pitting no matter how small that actual probability.
Further, we believe that single-population models lack physical
Factor Type Levels Values basis in this context, and our investigation has shown that other
Oxygen Fixed 2 Deaerated, aerated Table 5 EP NiTi—Fisher
s exact test indicates probability
Time Fixed 3 1, 12, 60 h of breakdown does not vary significantly with oxygen
concentration or duration of immersion time in PBS
Analysis of variance for Er using adjusted SS for tests
Oxygen Immersion Number Number Percent
Source DF Seq SS Adj. SS Adj. MS F P concentration time, h of tests pitted pitted

Oxygen 1 0.531 0.230 0.230 45.55 0.000 Deaerated 1 97 17 18

Time 2 0.106 0.0974 0.0487 9.65 0.000 12 56 4 7
Oxygen * Time 2 0.00024 0.00024 0.00012 0.02 0.976 60 12 0 0
Error 284 1.434 1.434 0.00505 Aerated 1 57 10 18
Total 289 2.071 12 55 7 13
S = 0.0710 R2 = 30.8% 2
R (adj) = 29.6% 60 6 1 17
Italics indicate model terms that are statistically significant at 0.05 Fisher
s exact test: P values of 1.0 and 0.265 for oxygen concentration
level and time, respectively

Table 6 EP NiTi—binary logistic regression confirms probability of breakdown does not vary significantly with oxygen
concentration or duration of immersion time in PBS
Binary logistic regression: censored vs. oxygen, time

Variable Value Count

Censored 1 252 (event)

0 39
Total 291

Logistic regression table

Predictor Coefficient SE coefficient Z P Odds ratio 95% CI lower 95% CI upper

Constant 1.40 0.296 4.73 0.000

Aeration
0.282 0.352
0.80 0.423 0.75 0.38 1.5
Time
12 h 0.712 0.385 1.85 0.064 2.04 0.96 4.33
60 h 1.305 1.05 1.24 0.214 3.69 0.47 28.9

Log-likelihood =
111.9. Test that all slopes are zero: G = 5.41, DF = 3, P value = 0.144

Journal of Materials Engineering and Performance

Fig. 11 Censored treatment of Eb
Er data from EP samples immersed in deaerated (left-two-parameter exponential) and aerated (censored
three-parameter Weibull distribution) PBS for 1, 12, and 60 h

probability distributions provide an adequate fit to the observed Other choices for F(Æ) can be considered; for example, using
Eb
Er data. the three-parameter Weibull distribution,
To overcome some of the limitations of the parametric
approach we have previously proposed (Ref 5), we derive a P2 ¼ 1
exp½
ð
c=aÞb ;
physically based, mixture model, using a Poisson distribution to where a and b are the (positive) scale and shape parameters,
describe the fraction of samples that will likely break down. respectively, and c is the (negative) threshold parameter.
This was done by modeling the occurrence of surface flaws, Alternatively, one could develop probability models for Eb
and then using a normal distribution to model the distribution and Er separately. Although we do not pursue such an
of breakdown potentials for flawed specimens. The Poisson approach in this paper, some justification can be found in the
distribution has been used previously to characterize the observed lack of correlation (r = 0.024, P = 0.882) between
number of pits expected (Ref 18). The Poisson probability the Eb and Er values for pitted specimens.
function ðpðkÞ ¼ exp
k kk =k!Þ actually gives the probability of The probability (P) that Eb
Er is negative can be
finding a specimen with k flaws that will pit before 0.8 V is approximated by:
reached. In computing this probability, k ¼ ha is proportional
to the area (a) of the surface being evaluated (1.15 cm2) and h P  PE ¼ P1  P2
is the expected number of occurrences per unit area.
where P1 is the observed proportion of breakdowns and P2 is
Our proposed approach to an acceptance criterion is to
computed by substituting data-derived estimates of l and r
estimate the probability of corrosion (P) based on the
in the expression for P2, as shown in Fig. 12 for the deaer-
probability that the sample will pit (P1) and, if the specimen
ated exposures and Fig. 13 for the aerated exposures. The
will pit, the probability (P2) that Eb
Er is negative:
next step is then to determine the upper bound PU for P. The
P ¼ P1  P2 ; lot of material can then be accepted or rejected if PU is less
than a specified limit on the probability for pitting and subse-
where P1 ¼ 1
expð
kÞ is the probability that any individ- quent breakdown.
ual sample will possess one or more flaws, and will therefore For our data, 39 out of 290 specimens broke down. With k
pit, and P2 = F(0), where F(Æ) is the cumulative distribution representing the expected number of flaws per 1.15 cm2 of
function that describes the variation in Eb
Er. We consider surface area, this rate can be estimated by
alternative parametric choices for F(Æ) by considering their
suitability for and goodness-of-fit to the values of Eb
Er k ¼
lnð1
P1 Þ ¼
lnð1
39=290Þ ¼ 0:144
observed for the pitted specimens. Using the normal distribu- or about one flaw per seven specimens.
tion for F(Æ), Restricting attention to the 39 pitted specimens, we can use


 analysis of variance to see if Eb is affected by oxygen
1 d
l
P2 ¼ 1 þ Erf pffiffiffi concentration or duration of exposure to PBS. The result of this
2 r 2
analysis is shown in Table 7 and confirms that Eb is not
is the probability that Eb
Er < 0 if the specimen pits. Here, significantly affected by oxygen concentration but is influenced
d is an arbitrary extra margin of safety above Eb
Er which by the duration of exposure. Similarly, an analysis of variance
for this analysis we have assumed to be zero. of Eb
Er (see Table 8) indicates that the margin of safety

Journal of Materials Engineering and Performance

Fig. 12 EP mixed probability model for deaerated PBS

Fig. 13 EP mixed probability model for aerated PBS

Table 7 Analysis of variance of Eb data from EP Table 8 Analysis of variance of Eb 2 Er data from EP
samples samples

General linear model: Eb vs. oxygen concentration and time General linear model: Eb 2 Er vs. oxygen concentration and time

Factor Type Levels Values Factor Type Levels Values

Oxygen Fixed 2 Deaerated, aerated Oxygen Fixed 2 Deaerated, aerated

Time Fixed 3 1, 12, 60 h Time Fixed 3 1, 12, 60 h

Analysis of variance for Eb using adjusted SS for tests Analysis of variance for Eb 2 Er using adjusted SS for tests

Source DF Seq SS Adj. SS Adj. MS F P Source DF Seq SS Adj. SS Adj. MS F P

Oxygen 1 0.0381 0.131 0.131 3.04 0.090 Oxygen 1 0.169 0.280 0.280 5.72 0.022
Time 2 0.437 0.437 0.219 5.10 0.011 Time 2 0.315 0.315 0.158 3.21 0.052
Error 35 1.50 1.50 0.0429 Error 35 1.72 1.72 0.0490
Total 38 1.98 Total 38 2.20
S = 0.207 R2 = 24.1% R2 (adj) = 17.6% S = 0.221 R2 = 22.0% R2 (adj) = 15.3%

Italics indicate model terms that are statistically significant at 0.05 level Italics indicate model terms that are statistically significant at 0.05 level

Journal of Materials Engineering and Performance

against pitting Eb
Er is not significantly affected by duration cyclic polarization testing using the Er values at the end
of exposure but is significantly affected by the dissolved of the pre-test immersion and the Eb from the testing in
oxygen concentration in the PBS. Thus, from this analysis we deaerated PBS.
see that for EP samples: 2. Acceptance criterion should be based on a calculation of
the probability that the device will pit in vivo is suffi-
1. Er is greater for aerated solutions ciently small. This can be done by cyclic polarization
2. Eb is similar for aerated and deaerated solutions testing a sufficient number of devices that have been pre-
3. Both Eb and Er increase with exposure time exposed to aerated PBS for sufficient times. This data
4. Eb
Er is lower in aerated solutions should then be used to calculate the probability of pitting
5. Eb
Er does not vary significantly with immersion time in vivo, i.e. the probability that Eb
Er is negative.
for aerated or deaerated solutions 3. The margin of safety (Eb
Er) of MP NiTi decreased sig-
nificantly after exposure to deaerated PBS at 37 °C.
Because aeration significantly affected Eb
Er and immer- Modeling showed that with a pre-test exposure of 1 h Eb
sion time did not, estimates of P2 are developed by fitting Er decreased approximately 296 mV when exposed to
probability distributions to the 21 pitted specimens tested under aerated rather than deaerated solutions. This change is
deaerated conditions and to the 18 pitted specimens tested primarily a result of the increase in Er rather than from a
under aerated conditions. Compared to a single-population change in Eb. If testing is done with pre-exposure to
approach, analysis using this physically based mixture model aerated solutions, the inherent variability in Eb
Er is
gives a more realistic result of 1 in 1600 for the probability of decreased and there is no need to account for the time
pitting for EP exposed to aerated PBS versus 1 in 730,000 for the implant will be in service.
specimens exposed to deaerated solutions. This can be 4. To achieve a high degree of confidence that the probabil-
compared to the probability for pitting of MP samples of 1 in ity of in vivo pitting is acceptably small, the amount of
1500 in deaerated PBS (60-156 h) and 1 in 1,800,000 for MP testing required will depend on the population variation
exposed to aerated PBS (1-156 h). in breakdown and rest potentials. Small-scale pilot testing
It appears that the MP samples have a very high surface should be performed when this information cannot be
density of pit-initiation sites and that the breakdown potentials obtained from other sources.
for these sites are reasonably well characterized by a normal 5. The pitting resistance (Eb
Er) of EP NiTi is higher than
distribution of voltages with a small standard deviation. The MP NiTi. Only five of the EP NiTi samples broke down
breakdown is most likely associated with defects in the oxide whereas all of the MP samples broke down. Using only
coating from mechanical polishing. The EP samples, however, the samples that broke down, the average Eb
Er was
appear to have a relatively low surface-defect density that 0.774 V SCE for the EP NiTi and 0.586 V SCE for the
results in pitting only once in 10 specimens. For the samples MP NiTi. This, however, does not take into account the
that pit, the distribution of the breakdown potentials are slightly fact that over 90% of the EP samples did not break
higher than those observed for the MP samples. It is possible down.
that once the defective oxide coating has been removed by 6. MP NiTi breakdown potentials can be treated with a nor-
electropolishing, the only defects that remain are where mal quality-assurance type of statistics and sampling the-
inclusions and cracks surrounding the inclusions intersect the ory. Eb
Er distributions can be used to predict the
surface. Shabalovskaya et al. investigated typical inclusions in probability that Eb
Er from the sample will be negative
NiTi and found that the inclusions are generally titanium rich as or some pre-determined threshold value. The use of a
we observed (Ref 21, 22). They indicate that these titanium rich threshold value can be used to account for the effect of
inclusions are generally Ti2NiO, TiO2, TiC, or oxidized TiC. drift in Eb and Er with long-term immersion, or testing can
The probability of inclusions, associated crevices, and voids be done on samples subjected to long-term immersion in
intersecting the surface is, hopefully, a rare event and may aerated solutions to verify that Eb
Er has stabilized.
explain why only one in seven of the EP samples failed. The 7. Quality-assurance testing and statistical analysis of EP
intersection of inclusions and voids and cracks surrounding NiTi, and other medical-device materials, with break-
such inclusions has been used in the past to characterize the down potentials frequently greater than 0.8 V SCE, is
probability of fatigue failure (Ref 23) and we believe a similar much more challenging. Here we propose a combined
process can be used to explain the pitting probabilities for NiTi model in which the censored data (no break down data)
once the more damaging/frequent oxide defects have been is used to model the probability that any given sample
removed by electropolishing. may, or may not break down as a Poisson process, then
model the breakdown potentials of the samples that have
failed as a normal distribution.
8. The probability that NiTi will pit in vivo (Eb
Er < 0)
7. Conclusion was estimated as:

1. Eb
Er as determined from the current version of ASTM (a) 1 in 15,500 for MP exposed to deaerated PBS for 1
F 2129 which does not include pre-test exposures to aer- and 12 h (Gumbel)
ated solutions and can over estimate the margin of safety (b) 0 for MP exposed to aerated for 1-156 h (three-
of NiTi to pitting corrosion in vivo. This indicates that parameter Weibull)
current acceptance testing of medical device implants (c) 1 in 730,000 for EP exposed to deaerated for 1 h
(such as ASTM F 2129) should be modified or extended (mixture model)
to include at least several days pre-test exposure of im- (d) 1 in 1,600 for EP exposed to aerated solutions for
plant material to aerated physiological solutions prior to 1 h (mixture model)

Journal of Materials Engineering and Performance

9. The estimated probability of pitting in vivo is higher for 10. G. Frankel, Pitting Corrosion, ASM Handbook, Vol 13A, S.D. Cramer
EP samples pre-exposed to aerated PBS than for deaer- and B.S. Covino Jr., Eds. ASM International, Materials Park, OH,
2003, p 236
ated PBS primarily as a result of the increase in dis-
11. J.R. Scully and M.J. Kelly, Methods for Determining Aqueous
solved oxygen raising Er. Corrosion Reaction Rates, ASM Handbook, Vol 13A, S.D. Cramer
and B.S. Covino Jr., Eds., ASM International, Materials Park, OH,
2003, p 68
12. D. Silverman, ‘‘Tutorial on Cyclic Potentiodynamic Polarization
Technique,’’ Paper no 299, CORROSION/98, NACE International,
References Houston, TX, 1998
13. D. Silverman, Practical Corrosion Prediction Using Electrochemical
1. B. Clarke, et al., Influence of Nitinol Wire Surface Treatment on Oxide
Techniques, in Uhlig
s Corrosion Handbook, 2nd ed., R.W. Revie, Ed.,
Thickness and Composition and Its Subsequent Effect on Corrosion
Wiley, New York, 2000, p 1179–1225
Resistance and Nickel Ion Release, J. Biomed. Mater. Res. A, 2006,
79(1), p 61–70 14. ‘‘Standard Specification for Wrought Nickel-Titanium Shape Memory
Alloys for Medical Devices and Surgical Implants,’’ ASTM F 2063-05,
2. L.E. Eiselstein, D.M. Proctor, and T.C. Flowers, Trivalent and
Vol. 13.01, ASTM International, West Conshohocken, PA, 2005
Hexavalent Chromium Issues in Medical Implants, Mater. Sci. Forum,
2007, 539–543, p 698–703 15. B.J. Burrow and J. Moskito, ‘‘Auger Electron Spectroscopy (AES)
Surface Analysis Report,’’ Job Number: C07S6006, Evans Analytical
3. J.J. Jacobs, J.L. Gilbert, and R.M. Urban, Current Concepts Review -
Group, Sunnyvale, CA 94086, USA, 2007
Corrosion of Metal Orthopaedic Implants, J. Bone Joint Surg., 1998,
80, p 268–282 16. ‘‘Implants for Surgery—Measurements of Open-Circuit Potential to
Assess Corrosion Behaviour of Metallic Implantable Materials and
4. ‘‘Standard Test Method for Conducting Cyclic Potentiodynamic
Medical Devices Over Extended Time Periods,’’ ISO 16429-04,
Polarization Measurements to Determine the Corrosion Susceptibility
Geneva, Switzerland. 2004
of Small Implant Devices,’’ ASTM F 2129-06, ASTM International
Standard F 2129-06, Vol. 13.01, ASTM International, West Cons- 17. T. Shibata and W.R. Whitney, Award Lecture: Statistical and Stochastic
hohocken, PA, 2006 Approaches to Localized Corrosio, Corrosion, 1996, 52(11), p 813–830
5. L. Eiselstein, et al., Toward an Acceptance Criterion for the Corrosion 18. M. Kowaka, Introduction to Life Prediction of Industrial Plant
Resistance of Medical Devices: A Statistical Study of the Pitting Materials, Allerton Press, Inc., New York, 1994
Susceptibility of Nitinol, Proceedings of SMST-2007 the International 19. T. Shibata, Corrosion Probability and Statistical Evaluation of Corro-
Conference on Shape Memory and Superelastic Technologies, 2007, sion Data, in Uhlig
s Corrosion Handbook, R.W. Revie, Ed., Wiley,
ASM International, Tsukuba City, Japan New York, 2000
6. R.A. Corbett, Laboratory Corrosion Testing of Medical Implants, 20. W.J. Conover, Practical Nonparametric Statistics, 2nd ed., Wiley Series
Proceedings of Materials and Processes for Medical Devices Confer- in Probability and Mathematical Statistics, Wiley, New York, 1980
ence, S. Shrivastava, Ed., ASM International, Materials Park, OH, 21. S. Shabalovskaya, J. Anderegg, and J.V. Humbeeck, Recent Observa-
2004, p 166-171 tions of Particulates in Nitinol, Mater. Sci. Eng. A, 2008, 481–482,
7. R.A. Corbett and S.N. Rosenbloom, An Assessment of ASTM F 2129 p 431–436
Test Results Comparing Nitinol to Other Implant Alloys, SMST 2006: 22. S. Shabalovskaya, et al., The Effect of Surface Particulates on the
Proceedings of the International Conference on Shape Memory and Corrosion Resistance of Nitinol Wire, SMST-2003: The International
Superelastic Technologies, 2006 Conference on Shape Memory and Superelastic Technologies, 2003,
8. S.N. Rosenbloom and R. Corbett, ‘‘An Assessment of ASTM F 2129 p 399–408
Electrochemical Testing of Small Medical Implants—Lessons 23. J. Schaffer, A Probabilistic Approach to Modeling Microstructural
Learned,’’ Paper no. 07674, NACE 2007, Houston, TX, 2007 Variability and Fatigue Behavior in ASTM F562 Medical Grade Wire,
9. B.G. Pound, Susceptibility of Nitinol to Localized Corrosion, Presented May 19th, 2006 at the 9th International Fatigue Congress,
J. Biomed. Mater. Res. A, 2006, 77(1), p 185–191 Atlanta, USA. 2006

Journal of Materials Engineering and Performance

View publication stats