ASTM F3211-17

This international standard was developed in accordance with internationally recognized principles on standardization established in the Decision on Principles for the
Development of International Standards, Guides and Recommendations issued by the World Trade Organization Technical Barriers to Trade (TBT) Committee.
Designation: F3211 − 17
Standard Guide for
Fatigue-to-Fracture (FtF) Methodology for Cardiovascular
Medical Devices
This standard is issued under the fixed designation F3211; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. A
superscript epsilon (´) indicates an editorial change since the last revision or reapproval.
1. Scope 2. Referenced Documents
2.1 ASTM Standards:
1.1 This guide is intended to provide an experimental
methodology to assess and determine the structural fatigue life E178 Practice for Dealing With Outlying Observations
E456 Terminology Relating to Quality and Statistics
of implantable cardiovascular medical devices.
E468 Practice for Presentation of Constant Amplitude Fa-
1.2 This guide is also intended to provide methodologies to
tigue Test Results for Metallic Materials
determine statistical bounds on fatigue life at in vivo use
E739 PracticeforStatisticalAnalysisofLinearorLinearized
conditions using measured fatigue life derived in whole or in
Stress-Life (S-N) and Strain-Life (ε-N) Fatigue Data
part from hyper-physiological testing to fracture.
E1823 TerminologyRelatingtoFatigueandFractureTesting
1.3 This guide may be used to assess or characterize device
F2477 Test Methods forin vitro Pulsatile Durability Testing
durability during design development and for testing to device
of Vascular Stents
product specifications.
F2942 Guide forin vitro Axial, Bending, and Torsional
Durability Testing of Vascular Stents
1.4 Fretting, wear, creep-fatigue, and absorbable materials
F3172 Guide for Design Verification Device Size and
are outside the scope of this guide, though elements of this
Sample Size Selection for Endovascular Devices
guide may be applicable.
2.2 ISO Standards:
1.5 As a guide, this document provides direction but does
ISO 5840-x Cardiovascular implants -- Cardiac valve pros-
not recommend a specific course of action. It is intended to
theses -- Part 1: General requirements, Part 2: Surgically
increase the awareness of information and approaches. This
implanted heart valve substitutes, Part 3: Heart valve
guide is not a test method. This guide does not establish a
substitutes implanted by transcatheter techniques
standard practice to follow in all cases.
ISO 12107 Metallic materials - Fatigue testing - Statistical
1.6 This guide is meant as a complement to other regulatory
planning and analysis of data
and device-specific guidance documents or standards and it
ISO 25539-x Cardiovascular implants -- Endovascular de-
does not supersede the recommendations or requirements of
vices -- Part 1: Endovascular prostheses, Part 2: Vascular
such documents.
stents, Part 3: Vena cava filters
1.7 This standard does not purport to address all of the
2.3 Regulatory Guidance:
safety concerns, if any, associated with its use. It is the
GuidanceforIndustry:Q9QualityRiskManagement,FDA,
responsibility of the user of this standard to establish appro-
priate safety, health and environmental practices and deter-
mine the applicability of regulatory limitations prior to use.
3. Terminology
1.8 This international standard was developed in accor-
3.1 Definitions of Terms Specific to This Standard:
dance with internationally recognized principles on standard-
3.1.1 acceptance criteria—specific numerical limits or
ization established in the Decision on Principles for the
ranges or other conditions identified prior to testing that
Development of International Standards, Guides and Recom-
mendations issued by the World Trade Organization Technical
Barriers to Trade (TBT) Committee.
For referenced ASTM standards, visit the ASTM website, www.astm.org, or
contact ASTM Customer Service at service@astm.org. For Annual Book of ASTM
Standards volume information, refer to the standard’s Document Summary page on
This test method is under the jurisdiction ofASTM Committee F04 on Medical the ASTM website.
and Surgical Materials and Devices and is the direct responsibility of Subcommittee Available from International Organization of Standards, http://www.ISO.org/
F04.30 on Cardiovascular Standards. ISO/store.htm
Current edition approved Sept. 1, 2017. Published September 2017. DOI: Accessed June 23, 2016 (http://www.fda.gov/downloads/Drugs/./Guidances/
10.1520/F3211-17. ucm073511.pdf).
Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. United States
F3211 − 17
establish the required results to support a conclusion, a 3.1.8 design life—thenumberofcyclesforwhichthedevice
decision, or meet a specification. is designed to remain functional without significant perfor-
mance degradation.
3.1.2 amplitude—one-half of the difference between the
3.1.9 device—a complete cardiovascular medical implant in
maximum and minimum measurements of the cyclic wave-
form. its final form, or as deployed, that may be used as a test
specimen.
3.1.3 censor—data where the cycle count at failure is only
3.1.10 duty cycle—a time history of loading conditions.
partially known. Run-outs (see definition in 3.1.26) are a form
EXAMPLE—For devices deployed into the vasculature of the
of right-censored data. Tests that use periodic inspections to
lower limbs, a duty cycle may be defined by the number of
determine the cycles to fracture are interval censored as the
steps per day, the number of stairs per day, and the number of
cycleoffractureisunknownbutboundedbetweentheprevious
sit/stand cycles per day.
and current inspection cycle counts.
3.1.11 failure—permanent deformation or fracture with
3.1.4 component—a test specimen comprised of a subas-
complete separation that renders the device ineffective or
semblyoranindividualpartofacardiovascularmedicaldevice
unable to adequately resist load. Other criteria may be used but
in its finished form.
should be clearly defined.
3.1.5 confidence level—the probability that the true value
3.1.12 failure mode—a combination of an external load
for a parameter of interest will fall within a numerical interval.
type, a fracture location or locations, and a fracture type. The
The interval is known as the Confidence Interval. Confidence
external load can be single modes such as bending or twisting
Intervals are used to establish boundaries for the value of a
torques, radial loads, tension-compression axial loads, and so
parameter of interest.
forth, or combinations of such loads. Fracture locations are
NOTE 1—Confidence levels, typically stated as percentages, are typi-
positions on a device at which fracture occurred such as in a
cally chosen through a risk analysis.
stent connector, stent apex, or stent strut. The fracture type is
3.1.6 coupon—a test specimen extracted from a cardiovas-
characterized by the surface morphology and the material
cular medical device or a component in its finished form.
cause or causes of the fracture such as tensile overload,
transverse shear, mixed-mode, high cycle fatigue, or low cycle
3.1.6.1 Discussion—Oftenacouponis“clipped”orcutfrom
fatigue.
an as-manufactured device.
3.1.7 design curve—the lower confidence bound for a reli-
3.1.13 fatigue factor of safety—the ratio of the Fatigue
abilityquantileofthefatiguelifedistribution.Forexample,the Strength at a Specified Life with prescribed reliability and
Load versus fatigue life Number of cycles (S-N) curve for p%
confidencelevelstotheloadatthespecifiedusecondition.The
survival at c% confidence. See Fig. 1. Fatigue Factor of Safety is specific to a single failure mode.
FIG. 1 Fatigue Life Model Depicting Terminology Where S is Load Parameter and N is Fatigue Life, Number of Cycles to Fracture
F3211 − 17
3.1.13.1 Discussion—When mean loads are considered on in vitro data and modeling. In general, higher reliability in
along with the alternating loads, the ratio calculation must be FtF is expected to increase the clinical reliability.
defined and preferably shown on a constant life fatigue
3.1.25 risk analysis—(1) a methodical analytical approach
diagram.
to determine and address identified system or component
3.1.13.2 Discussion—In communicating a Fatigue Factor of
failure modes and their associated causes, based on the
Safety, a clear statement of its intended purpose and the
probability of occurrence and the severity of their effects on
assumptions associated with its calculation is necessary for
system performance and patient safety; (2) an estimate of the
proper interpretation. For example, a safety factor estimate
riskassociatedwithidentifiedhazardsinaccordancewithFDA
based on the average amplitude at fracture at the design life
Q9 Quality Risk Management.
relative to the amplitude at the typical use condition will be
3.1.26 run-out—no fatigue failure at a specified number of
substantially different from a safety factor based on the 90 %
load cycles. See Terminology E1823. This number is typically
reliability/95 % confidence amplitude at fracture at the design
specified prior to beginning the testing.
life relative to a conservative estimate of the most challenging
3.1.27 sample size—the quantity of individual specimens
use condition amplitude.
tested. The sample size is typically chosen to establish confor-
3.1.14 fatigue life model—a mathematical equation that
mance to a pre-determined specification with appropriate
describes the relationship between fatigue life and loading
statistical confidence levels.
parameters with prescribed reliability and confidence, statisti-
3.1.28 load versus life (S-N) curve—graphical representa-
cally derived from experimental fatigue data. See Section 7.2 .
tion of fatigue life data (see Fig. 1). The curve indicates the
3.1.15 fatigue strength at a specified life—the maximum
load versus cycles-to-fracture relationship for a specified
th th th
load the test specimen can be expected to survive for a
probability of survival, for example, the 50 ,90 ,or95
specified number of cycles with a stated confidence and
percentile.
reliability.
NOTE 4—For N, a log scale is commonly used. For loads in stress or
3.1.15.1 Discussion—The Design Curve at a specified life
strain, either a logarithmic or a linear scale is commonly used. See
may be used to show this graphically. See Fig. 1.
Terminology E1823. For the purpose of analysis, the S-N curve is
commonly modeled using a load-life relationship, for example a Power
3.1.15.2 Discussion—The Fatigue Strength is specific to a
Law or Coffin-Manson equation.
single failure mode. See Terminology E1823.
3.1.29 strength distribution at life N—the probability of
3.1.16 fracture—complete separation of any device compo-
fractureatthelifeNasafunctionofload.Thedistributionmay
nent due to stress with exposure of new surfaces that were
be computed by integrating the fatigue life distribution at each
previously together.
load from 0 to N.
NOTE 2—A fracture does not necessarily represent a device functional
3.1.30 surrogate—atestspecimenconstructedtorepresenta
failure.
device, component, or region of interest of a cardiovascular
3.1.17 FtF—acronym for Fatigue-to-Fracture.
medical device in its finished form.
3.1.18 hyper-physiological test conditions—test loads that
3.1.31 test artifact—spurious test results attributable to
exceed the expected in vivo use conditions.
conditions that are not present during in vivo use conditions
(failure at the grips, for example).
3.1.19 load—used to denote continuous and time-varying
3.1.32 test specimen—a test article that is subjected to
forces, stresses, strains, torques, deflections, twists or other
parameters that describe the applied fatigue stimuli. Typically fatigue loading conditions.Atest specimen (also referred to as
specimen) may be classified as a device, component, coupon,
these fatigue stimuli are described by a mean value and an
alternating value. or surrogate.
3.1.33 test-to-success—a paradigm for assessing or charac-
NOTE 3—Units and symbols are dependent on the parameter of interest.
terizing the fatigue durability of medical devices whereby
3.1.20 physiological loads—loads expected on the device
specimens are tested at a chosen factor of safety at or near
during in vivo use.
simulated cyclic physiological loads where no fractures are
3.1.21 preconditioning—simulated use preparation of the
expected. For example, the device “passes” and the test is
specimen prior to testing. See Section 6.12.
successful if no devices fail by structural fracture or if all
devices maintain sufficient functional integrity. See Test Meth-
3.1.22 protocol—a set of instructions that typically defines
ods F2477.
the specimens, test procedures, analysis procedures, and ac-
3.1.34 use conditions—the conditions to which the device
ceptance criteria.
will be subject, including the cumulative effects of the final
3.1.23 quantile—value such that a fraction of the sample or
manufacturing state, the process of device delivery and
population is less than or equal to that value. See Terminology
deployment, and the in vivo operating environment. See 6.1
E456.
and 6.12.
3.1.24 reliability——the probability of survival to the speci-
4. Summary of Guide
fied design life at a given loading condition.
3.1.24.1 Discussion—For the purpose of this standard, this 4.1 The fatigue-to-fracture (FtF) paradigm provides a meth-
is a narrow statistical measure of reliability of the device based odology whereby whole devices, device components, coupons
F3211 − 17
or surrogates are tested to fracture with hyperphysiological 5.2.1 While the FtF methodology applies only to bench
cyclic mechanical loads such as deflections, forces, or torques. tests, it can provide insights into device behavior that would
In many or all of the tests, the cyclic load should be sufficient not necessarily be apparent in clinical studies that typically
to fracture the device in fewer cycles than the desired clinical focus on patient outcomes. After appropriate boundary condi-
life. The resulting fatigue data are used to make a statistical tions such as loadings, fixturing, and materials have been
estimate of fatigue life and/or generate outputs such as a determined, the FtF methodology can provide extensive infor-
fatigue safety factor and a fatigue strength distribution at the mation on the expected longevity of a device in a period 10 to
design life. 1000 times shorter than a real-time clinical study.
5.2.2 FtF is informative in characterizing device behavior
4.2 Thisdocumentprovidesguidancefortestconsiderations
over a wide range of loads and cycles. This is especially
and choices such as determining physiologically relevant test
valuable when the in vivo loading mode is understood but the
modes, determining load levels, selecting test specimens,
load magnitude and cycle requirements are not well known or
defining failure, characterizing and verifying test operation,
when characterizing device performance over a wide range of
selecting the test environment, determining an appropriate
patient lifetimes, activity levels, and physiological states is
sample size, setting the test frequency, setting the test duration,
desired.
preconditioning test specimens, monitoring the test, inspecting
5.2.3 InFtF,testloadsgreaterthanthedevices’expecteduse
for fractures, and documenting test results.
conditions are used. Thus, factors of safety can be measured
4.3 Prospective test planning procedures are illustrated to
relative to expected in vivo use conditions in both loading/
generate a credible estimate of durability relative to the in vivo
deformation severity and number of cycles.
useconditions.Theplanningprocedurecanbeusedtogenerate
5.2.4 In FtF, the nature and location of fractures observed as
a test protocol that includes a prospectively chosen statistical
a function of load can help provide insights into the device
model, sample size and test load levels, and rationale for the
response to the applied loading. The identified primary and
choices.
follow-on fracture locations and modes may be used to assess
the credibility of device computational models, as well as to
4.4 This document provides guidance on statistical interpre-
tation and presentation such as selecting the fatigue life model, evaluate potential impacts on clinical safety and efficacy,
especially post-fracture.
calculating confidence bounds, choosing between Frequentist
and Bayesian statistical procedures, and avoiding common 5.2.5 The FtF methodology can quickly and reliably assess
the impact of changes in processes, materials, or small changes
statistical pitfalls.
in geometry on in vitro fatigue life. These assessments with
respect to fracture can be quantified and used as part of
5. Significance and Use
validating design changes, demonstrating that the device meets
5.1 Use of this Methodology:
product specifications, or as part of guiding design improve-
5.1.1 This guide provides a compendium of information on
ments.
methods to use fracture data, fatigue life models, and statistical
5.2.6 FtF testing can often be completed in a shorter period
techniques to estimate the structural fatigue durability of an
of time than test-to-success testing since the FtF tests are
implantable medical device under anticipated in vivo loading
typicallyterminatedatasmallernumberofcycles.Specifically,
modes. The methodology for high-cycle fatigue assessment
whenextrapolationincyclesisappropriate,comparisonsofthe
relies on hyper-physiological tests intended to cause device
loads or the frequency of fracture at a lower number of cycles
fractures. Using the FtF methodology, fractures should not be
can provide a useful measure of equivalence.
avoided during testing; instead they provide the information
required to statistically assess device longevity under a wide
6. Procedure for Testing
variety of physiological and hyper-physiological test condi-
6.1 Determine Physiological Loads:
tions.
6.1.1 Since the FtF methodology is for bench testing, it is
5.1.2 Through evaluation of fracture locations, the geom-
essentialthatthefullrangeofclinicallyrelevantloadingmodes
etries after fractures, and the use conditions of the device, this
and magnitudes be identified or bounded. Guidance documents
guide may be used to help assess device safety.
from regulatory agencies such as the US FDA, guides and
5.1.3 This guide may be used to help assess differences in
standards from organizations such as ASTM or ISO, clinical
fatigue life between different devices or device histories. The
literature, and medical imaging and observations may provide
effects on fatigue life due to changes to a device’s geometry,
usefulrecommendationsonapplicabletypesandmagnitudesof
processing, or material may be assessed using this guide.
loads for device fatigue assessment.
5.1.4 Users of this guide must keep in mind that bench tests
6.1.2 For the intended patient population, the manufacturer
are simulations of in-use conditions. Adherence to this guide
should identify the use conditions, the design life, the potential
may not guarantee that results translate to individual clinical
of device fracture to produce adverse events, and the intended
scenarios. Therefore, in assessing a device’s fatigue
claims.Ifparticularpatientsub-populationspresentprocedural,
performance, the results from Fatigue to Fracture testing
operating, or lifetime conditions beyond the final product
should be reviewed in combination with other available data,
such as animal studies, clinical experience, and computational
simulations.
The term “manufacturer” is used in this guidance to mean “user of this
5.2 Significance of this Methodology: standard”.
F3211 − 17
requirement, a description of those conditions and a rationale bly that is practical.Test specimens should be representative of
for exclusion may be useful. The use of Design Failure Mode actual clinical components made by the final manufacturing
and EffectsAnalysis (DFMEA) and other risk analysis tools in process.
this identification process is encouraged (for example, see
6.4.2 For devices where a single size implant is used over a
Mikulak (1) or Teixeira (2)).
range of application sizes (vessel diameters, for example),
6.1.3 When determining in vivo use conditions, consider-
either assess the maximum and minimum use diameters or
ation should be given to the types, ranges, and duty cycles of
determine and assess the most challenging use condition based
conditions in the intended population. Estimates of in vivo
on stress analysis or experimental data. The manufacturer
fatigue life are strongly dependent on in vivo boundary
should take into account any interactions between the device
conditions that vary from patient to patient and activity to
and in vivouseconditionsovertherangeofapplicationsizesto
activity. Imaging or modeling the device’s or a well character-
determine the most challenged device size. See 6.8.4 for one
ized similar device’s deformations in vivo is encouraged.
example where vessel diameter may be important.
6.1.3.1 Limitations on the accuracy and generalizability of
6.4.3 If devices come in multiple sizes with a common
in vivo measurements should be noted and reported; for
design application, geometric architecture, and materials and
example, single-plane x-ray clinical measurements on sedated
processes,thenexperimentalorcomputationalmethodsmaybe
patients may not accurately represent the geometry or range of
used to determine the most challenged size and the FtF testing
actual physiological conditions.
may be confined to that size as a representation for the entire
6.2 Determine Durability Requirements: size range.
6.2.1 With the intended patient population in mind and the
6.4.4 Select test specimens that are representative of the
potential hazards associated with durability, the manufacturer
finished device. Consider test specimen features that may
shouldestablishtheclinicaldurabilityrequirementssuchasthe
influence the fatigue test results such as surface finish, micro-
device loads and/or the device deformations, the minimum
structure(grainsizeandtexture),loadingorientation,geometry
number of cycles to fracture or failure, and the failure criteria.
and dimensions, mechanical properties, cold work, residual
stresses, size and distribution of material or process flaws, and
6.3 Choose Test Modes:
preconditioning.Considerthesefeaturesandanyotherrelevant
6.3.1 The manufacturer should relate the selection of test
factors if specimens other than the finished device are to be
modes to the known and predicted interactions between the
justified for use in fatigue testing. If a coupon or surrogate is
implant site and the implanted device. Fatigue testing should
made to facilitate fatigue testing, a numerical model may also
be performed to elicit the anticipated in vivo mechanics; for
be used to demonstrate the similarity in stress distribution
example, a torsion fatigue test is not likely to be informative if
between the coupon or surrogate and the actual device under
in vivo bending fatigue is anticipated.
testing conditions.
6.3.2 If devices will be exposed to multiple modes of cyclic
physiological loads (such as radial compression, bending, 6.4.5 Though elements of FtF may be applicable, testing of
standard test specimens (ASTM “dogbones”, for example) or
torsion, flattening, axial tension/compression, and so forth),
consideration should be given to the effects of combined other ideal geometries is considered classical fatigue and can
loading. In each case, the manufacturer should relate the be planned and analyzed using classical methodologies such as
in ISO 12107 and Practice E739.
magnitudes of each mode tested and the manner in which the
loading is combined, or tested in isolation, to represent the in
6.4.6 Sample selection procedures should follow good sta-
vivo use conditions.
tistical practices to produce a representative sample (see ISO
6.3.3 If one mode clearly dominates the fatigue life, single-
12107). Randomization in sample selection, such as using a
modetestingtofracturetoestablishafactorofsafetycombined
random number generator, is recommended whenever practi-
with analysis of that mode plus the secondary modes may
cable to assure a high degree of independence in the contribu-
eliminate the need to test one or more of the secondary modes
tions of experimental error to estimates of treatment effects
in conjunction with the dominant mode. With appropriate
(see Terminology E456).
evidence, the manufacturer may choose to exclude loading
6.5 Define Failure in Fatigue:
modes that are not expected to result in fracture or loss of
6.5.1 A clear definition of the test’s acceptance criteria
function.
should be established. Typically this is chosen to be consistent
6.3.4 General guidance to some testing modes is given in
with the specific failure mode(s) identified by the FMEA or
Guide F2942, ISO 25539, and ISO 5840.
other risk analysis.
6.4 Select Test Specimens:
6.5.2 First fracture may be used as the definition of failure.
6.4.1 Test specimens should be nominal finished devices,
However, depending on the application and type of fracture,
appropriate components, coupons extracted from the device or
the specimen may still be functionally adequate with one or
component, or surrogate samples. In order to best reveal
more fractures.
unforeseen and characterize known failure modes, preference
6.5.3 If one or more fractures are acceptable within an
should be given to testing full devices or the largest subassem-
individual specimen, the manufacturer should define the crite-
ria and provide supporting evidence to distinguish acceptable
from unacceptable fracture(s). However, all acceptable and
The boldface numbers in parentheses refer to a list of references at the end of
this standard. unacceptable fractures should still be reported and summarized
F3211 − 17
in the test report. Recommended post-fracture test procedures of conditions that will induce both fractures and run-outs. The
are discussed in 6.11.6. allocation of specimens to hyper-physiological test conditions,
6.5.4 Prescribed acceptance criteria may be established to where fractures are expected, and Test-to-Success conditions,
excludeoccurrencesoffracturethatareartifactsofthetest(see where no fractures are expected, will depend on considerations
6.15.6). of the FMEAor other risk analysis, the raw material behavior,
6.5.5 It may be appropriate to use a failure criterion defined theuseenvironment,andthestatisticalmodelstobeemployed.
by the loss of acceptable function such as mechanical perfor-
6.8.2 The fluctuating loads on a device or test specimen
mance without an actual fracture taking place. For example, which induce fatigue can vary in type, magnitude, and fre-
cyclic stress-softening could reduce mechanical stiffness to an
quency. For S-N characterization, typically employed in clas-
unacceptable level, or cumulative plastic damage could reduce sical fatigue and FtF, constant-type/constant-magnitude/
the diameter to an unacceptable level. If FtF is used in these
constant-frequency cyclic tests are used (for example see ISO
circumstances, the test report should address whether or not 12107 or Practice E739). For a given test frequency, these test
such device behavior is expected, how it is accounted for, and
conditions can be described by two parameters: the load
how functional failure is determined and statistically analyzed. amplitude and the mean load. Often, the load level combina-
tions are created by increasing the load amplitude while
6.6 Characterize the Test:
keeping the mean load constant, the ratio of (load amplitude)/
6.6.1 The manufacturer should assess the impact of ideal-
(mean load) constant, or the ratio of (minimum load)/
izations and simplifications present in the test setup, operation
(maximumload)constant.Recordthemethodologychosenand
and test specimens that may impact the results, such as: the
7 provide a scientific rationale for its use such as feasibility test
boundary conditions , machine alignment, machine stability
data or historical experience.
and durability, preconditioning, device alignment, device
6.8.3 Given previous material/device fatigue
orientation, device positioning, device non-uniformity, post-
characterization, one load level may be sufficient to compare
fracture behavior, and any pre- and/or post-deployment proce-
FtF testing of two similar designs. Also, one load level is
dures. The assessment may be used to determine what factors
typically used in the Test-to-Success approach to demonstrate
need to be controlled in the test.
no or few failures. When less historical data are available, a
NOTE 5—The following characterization activities may be useful:
minimum of two levels are required to demonstrate a transition
• Observe the geometry and displacements over the range of test
from majority-fracture condition to majority-run-out condi-
frequencies and amplitudes using high-speed video and image analysis
tions. When little pre-existing data are available, or to truly
software(ifappropriate).Observethewholespecimen,payingattentionto
potential fracture locations and the apposition between the test specimen define a transition in regions, such as a plateau from low-cycle
and the fixtures used to impose the boundary conditions. When load or
to high-cycle fatigue (see Dowling (3) for examples), a
strain-rate sensitive fixtures are used, the cycle rate should be sufficiently
minimum of three levels is required. In this case, typically two
slow to ensure that the specimens maintain continual apposition to the
levels are in the shorter life domain to establish an S-N slope
testing fixtures.
for that region and the third level in the majority run-out
• If test specimens vary substantially in size, stiffness, mass, or other
design attributes, it is desirable to observe test operation over the full
condition to establish a change in slope. If a fourth test
range of specimens.
condition, equivalent to Test-to-Success load and cycle life
• Observe and characterize variations in the applied load (for example
condition is tested, or if the run-out cycle number with
displacement, curvature, forces, torques, and so forth) over an appropriate
super-physiological loads exceeds the design life, then no
sampling period at various intervals from the beginning to the end of
extrapolation would be required.
testing.Rapidchangesinloadsmaybeindicativeofafractureorachange
in deformation mode.
6.8.4 Thelevelsshouldbechosentoincorporatevariationin
• Force measurements during setup in a deformation controlled test
the amplitudes, the mean, or both. The effect of mean load on
can be useful in assuring that the intended conditions are being imposed
the fatigue life may differ between low-cycle fatigue and
on the test specimens.
high-cycle fatigue, between force-control and deformation-
• Assessthepotentialfortestartifactstoinducefracture,suchasstress
concentrations associated with rigid grips. control fatigue tests (Manson (4)), and between materials with
• Assess the potential for force and displacement measurement errors
residual stresses and those without residual stresses.
as a result of specimen and fixture geometric tolerances, signal collection
and filtering, inertial and frictional effects, and so forth. NOTE 6—On specific cases:
• In low-cycle deformation-controlled fatigue of devices with materials
6.7 Verify Test Operation:
whose stresses reduce or accumulate plastic damage with cyclic use,
6.7.1 Through dimensional measurements, video and still
initial mean test conditions may have little influence on fatigue life.
imagery, strain gages, modeling, and any other appropriate • In high-cycle fatigue, in either force- or deformation-control, usually
thereislittleplasticityandcorrespondinglynoorlittlechangeinthemean
characterization technique, show that the devices are deform-
load.Thusmeanloadstendtoinfluencehigh-cyclefatiguelife.Theremay
ing in the intended manner, the loads are as expected, and the
be most-challenging mean conditions; for example in radial fatigue of
counts of cycles are accurate.
some self-expanding stents the mean strains are higher in small diameter
vessels than in large diameter vessels.
6.8 Select Test Conditions:
• In load control, devices with stress hardening materials may plasti-
6.8.1 In general, the in vitro fatigue properties of a device
cally deform initially, but then stabilize and have good fatigue life.
willbemosteffectivelycharacterizedbytestingunderavariety
However, devices with strain-softening materials may fail quickly if the
deformation magnitude increases from cycle to cycle.
• The presence of residual stress in low-yield stress materials tested
Boundary Conditions refers to loading or deformation conditions imposed on
the test sample, geometric constraints that control the force and moment, or under low-cycle fatigue conditions tends to have minimal impact on
deformations at locations where the test article interacts with the testing fixtures. fatigue life because cyclic plasticity will tend to reduce the levels of
F3211 − 17
residual stress. Conversely, the presence of residual stresses in high-yield
linear or linearized S-N data and testing all specimens to
stress materials tested under high-cycle fatigue conditions tends to change
fracture, Practice E739 provides useful practice on the mini-
the mean life and increase the variation in fatigue life. See Withers (5).
mum sample size and load levels. For testing to product
6.8.5 In FtF feasibility testing to find load levels for
specifications, the number of load levels and the sample size
transition from fracture to run-out, studies are typically run at
should be justified by the manufacturer, typically based on the
multiple levels with relatively large changes in load levels and
FMEA or other risk analysis and any preexisting fatigue data.
small numbers of samples. It may be efficient to start at very
6.9.2 For product specification testing devices with a range
highlevelsthatinducefracturequickly,showthefailuremodes
of different sizes, Guide F3172 provides guidance for selecting
and perhaps show changes in failure modes. Alternatively for
the appropriate size(s) and determining an appropriate number
feasibility studies, especially if specimens or setups are
of samples. Though that guidance assumes the testing is at a
expensive, starting at low levels, continuing to run-out, and
single-load level, the concepts can be extended to testing at
reusingthenonfracturedspecimensafterraisingthelevelsmay
multiple levels.
also be efficient, though one has to be aware of the potential of
coaxing as well as effects such as cumulative damage and 6.9.3 When testing a single specimen type to estimate the
cyclic hardening or softening. Reusing run-out specimens
median load to fracture at a given number of cycles, an
should be avoided in product specification testing.
“up-down” or “staircase” method is efficient in the use of
6.8.6 The lowest FtF test conditions will usually be at
samples.The basic up-down technique is to start at an estimate
“challenging”or“extreme”physiologicalconditions.Itmaybe
of the load associated with median fatigue strength for the
useful to choose levels via hyper-physiological modeling, for
desired number of cycles. If the specimen does not fracture by
example, raise the load levels from normotensive systolic/
run-out at that load, then run the next test sample at one load
diastolic pressures of 120/80 mmHg to severely hypertensive
increment (mean, amplitude or both) greater in load.
levels such as 180/110 mmHg (see ISO 5840-1) and above.
Otherwise, test at one load increment lower. The load incre-
6.8.7 The highest test conditions should be below where the
ment is typically chosen in the range of 0.5 to 1.5 times the
applied mechanical loads cause non-clinically relevant failure
standard deviation of the underlying distribution (see Little
modes. Deformation observation, load deflection curves,
(6)).Theaccuracyandspeedofthisapproachdependsuponthe
and/or computational stress-strain analysis may be useful to
size of the load increment, the initial load level, the repeatabil-
predict forces or amplitudes at which failure modes become
ity of the load application, the sample consistency, and the
non-relevant to the clinical situation, such as local buckling or
number of test specimens. The number of samples required to
local macroscopic plastic yielding.
obtain a given level of confidence can be determined using
6.8.8 To determine the slope of the S-N curve for a given
standard statistical techniques (ISO 12107).
sample size, the difference in load levels tested should be
proportional to slope; for example, if the life increases rapidly 6.9.4 Alternatively, an optimal sampling method can be
with changes in load levels in some region, use small changes used to provide load levels to obtain the median fatigue
in the load level to measure the slope in that region.
strength and associated standard deviation (see Neyer (7)).
6.8.9 With increases in variability in nominally equivalent
This approach improves upon traditional up-down testing by
test specimens or variability in setup or operation of the test
optimizing the choice of load levels based on previous testing
apparatus, increasing numbers of samples are required to
results.An estimate of median fatigue strength and its standard
accurately determine the slope and distribution of the S-N
deviation are updated sequentially and testing is continued
curve.
until the desired confidence is reached (see Fig. 2).
6.8.10 Tests at different load levels may have different
6.9.5 Feasibility or historical data can be used to justify and
fracture modes. In order for any extrapolation to be valid, the
develop test plans using statistical re-sampling; i.e., the cre-
fracture mode should remain the same from the data domain to
ation of statistically valid, but fictional, data. The following
the extrapolated domain. Thus, the failure mode or modes in
iterative procedure may be utilized to choose a test plan:
the extrapolated domain of interest should be determined and
(1) Fit a Fatigue Life Distribution to a feasibility or
thepredictionbasedononlythosefracturesinthedatadomain.
historical dataset using one of the statistical analysis proce-
See 6.15.
dures described in Section 7 (see also references in 7.2.2,
6.9 Determine the Sample Size and Plan the Test:
especially NIST/SEMATECH (8)).
6.9.1 The number of specimens and load levels required in
(2) Propose a Test Plan.ATest Plan consists of a specified
fatigue testing depends on the type of test program (prelimi-
setofloadlevels,thenumberofsamplesateachloadlevel,and
nary and exploratory, research and development, or product
the test duration.
performance tests, for example). For some test programs with
8 10
Coaxing is an effect that can occur in some materials where the high-cycle It is recommended the data represents two or more load levels that show a
fatigue strength increases when test articles are cyclically preconditioned at lower majority of fractures within or at higher cycle counts than the knee (transition from
stresses (or strains). low- to high-cycle fatigue). The Knee is a point on a load-life curve at which the
Clinically relevant failure mode is the manner in which a failure may slope changes, separating the region of low cycle fatigue where plastic yielding
potentiallyoccur in vivothatcouldcauseharmorreducethesafetyorefficacyofthe dominates and the region of fatigue where the macroscopic response is elastic. The
delivery or function of the device. Clinically relevant failure modes can be kneemaynotbedistinct.Ifchoosingakneepointforpredictivemodeling,statehow
established through clinical evidence, experimentation or simulation. the point was chosen.
F3211 − 17
FIG. 2 History Plot Showing Sequence of Stimuli Levels Tested by way of Optimal Sampling Method
NOTE 1—The symbol x denotes a fracture prior to run-out and o denotes a run-out. Sidebar statistical analysis shows the predicted cumulative
probability of fracture curve (green) and the fatigue strength distribution at the specified life of 1 million cycles (red). The fatigue strength at the median
life is marked (green).
(3) Generate statistically valid but fake data using the fatigue testing of NiTi specimens if it suppresses the stress-
Fatigue Life Distribution and the Test Plan. Analyze the fake induced phase transformation.
dataasifitweretheoutcomeofthetest,i.e.useittodetermine
6.10.1.3 High test frequency may reduce the effects of the
the results of interest (fatigue strength or factor of safety, for
environment on fatigue performance.
example (see 7.5)).
6.11 Determine Test Duration (Run-out) & Stopping Crite-
(4) Generate enough simulated data to assess whether or
ria:
not the Test Plan has a good chance to differentiate between
6.11.1 The test may be stopped for the entire cohort when
pass and fail at the design life for the targeted device
one, some, or all samples have fractured, failed, or the cycle
performance level.
count reaches a predetermined number of cycles (in other
(5) Repeat steps (2), (3), and (4) in order to refine the test
words, run-out). Similarly, the testing of individual specimens
plan by repeatedly resampling the simulated test results.
may be stopped at fracture, failure, or run-out. The criteria for
6.10 Determine Test Frequency:
discontinuing testing on both the cohort and individual speci-
6.10.1 Cycle rates used for durability testing should be
men level should be included in the test protocol.
justified based on factors such as the material’s strain-rate
6.11.2 When using the FtF paradigm, especially during
dependency, temperature sensitivity, and environmental
design development, the number of cycles selected for run-out
resistance, the device-test apparatus rate-dependent dynamic
may be less than the design life of the device. However, if the
loads, and the thermal equilibrium conditions.
material, test method and fatigue life model do not have an
6.10.1.1 For materials that are deformation rate-dependent,
established historical or scientific basis, restricting all the
the testing frequency may be chosen at or near the in vivo
testingtoshorterthanthedesignlifeshouldbeavoidedtolimit
frequency. Faster rates are acceptable if evidence and scientific
the uncertainty in extrapolation. Since the slope of the fatigue
justification are provided.
lifesurvivalcurvemaychange,testloadsshouldbeselectedso
6.10.1.2 For materials with mechanical hysteresis loops and
that a sufficient quantity of the testing is in the domain of
temperature sensitivity, strain rates should be slow enough that
interest to increase confidence in extrapolation to the design
the test environment (air, water, saline, for example) can
life domain. Feasibility testing, as recommended in 6.9, should
maintain the entire cross-section of the device near the target
be used to determine these load levels.
temperature. For example, for Nitinol (NiTi) components with
smallcross-sections,testspeedaccelerationisusuallypossible, 6.11.3 In order to avoid any extrapolation to larger numbers
since heat diffusion rates in water or saline are typically fast of cycles, the number of cycles at run-out may be chosen to
enough to maintain the material near the saline temperature equal the design life of the implant. This provides direct
(see Yin (9)). Excessive test speed may be detrimental to the evidence of sufficiency of the product at the design life at the
F3211 − 17
test conditions, which may be at or above design use condi- 6.12.2 Depending on the test objectives, preconditioning
tions. Fatigue-to-fracture testing of this type is recommended may also include packaging distribution, sterilization, and
storage simulations.
when determining both fracture and non-fracture load condi-
tions at the design life.
6.13 Monitor Test Setup and Operation:
6.11.4 If a specimen is found to fracture after the run-out
6.13.1 In the test setup for deformation-controlled tests,
criterion, the cycle count at the prior inspection interval should
measurement of the peak force at the minimum and maximum
be recorded. With automatic fracture detection, the cycle at
deflection is recommended to assure that the setup is suffi-
fracture would be recorded. For most statistical purpo
...