Ankle arthritis is a degenerative condition that can cause substantial pain
and functional limitation. Treatment options for symptomatic ankle arthritis
include anti-inflammatory medications, orthotic devices, and operative
débridement. End-stage ankle arthritis refractory to these treatments
has traditionally been treated with ankle fusion. Successful ankle fusion
relieves pain but reduces the range of motion, alters gait kinematics, and may
increase the risk of arthritis in adjacent
joints1-4.
Total ankle arthroplasty has been utilized in the past to overcome the
limitations of ankle fusion, but early implant designs resulted in high rates
of
failure1,5-10.
Recently, several new implant designs have been introduced, and studies have
demonstrated promising short and medium-term results with the use of those
devices. On the basis of these reports, there has been a renewed interest in
total ankle arthroplasty as an alternative to ankle
fusion11-20.
Despite a lack of long-term clinical data, the use of total ankle
arthroplasty is expanding. The purpose of this cost-effectiveness analysis was
to assist orthopaedic surgeons in evaluating whether the currently available
literature supports the emerging use of total ankle arthroplasty at this time.
In this study, we also extrapolated published data on the survival and
function of ankle prostheses to identify threshold values that would better
justify the use of total ankle replacement. The actual cost-effectiveness of
total ankle arthroplasty will not be known until longer-term data on the new
prosthetic designs are published. The thresholds identified in this study can
help orthopaedic surgeons to evaluate the results of clinical trials as they
become available and to decide whether to perform total ankle
arthroplasty.
General Framing and Design
This cost-effectiveness analysis followed the methodological guidelines of
the Panel on Cost-Effectiveness in Health and Medicine convened by the United
States Public Health Service in
199321. The panel
outlined an explicit set of recommendations in a reference case analysis.
These reference case guidelines established a common set of standards in order
to improve the comparability of cost-effectiveness analyses. Issues addressed
in the reference case analysis include standard practices for framing and
perspective of the study, identification of outcomes, estimation of costs, and
testing of
uncertainty21. This
study was constructed with adherence to those standards.
Consistent with the reference case guidelines, this analysis compared the
cost-effectiveness of total ankle replacement with that of ankle fusion, the
accepted treatment of ankle arthritis refractory to nonoperative
management1. The
analysis assumed a target patient population with end-stage ankle
osteoarthritis at the age of fifty-five years. The time horizon of the
analysis encompassed the remaining life expectancy of this target
population.
The cost-effectiveness ratios for total ankle replacement and ankle fusion
were analyzed from a societal perspective. The boundary of the analysis is
limited to the costs and health effects directly impacting the target
population. Costs, effectiveness, and the probability of various outcomes were
estimated on the basis of data obtained from a literature review.
Decision Model
Decision tree software (DATA 4.0; TreeAge Software, William-stown,
Massachusetts, 2001) was used to create a model for the treatment of end-stage
ankle
arthritis21-29.
The decision tree is shown in Figure
1. The model begins with the decision to perform either ankle
fusion or total ankle replacement. The literature was reviewed to identify
possible outcomes and their probabilities following each of these procedures.
These event pathways were incorporated as branches in the decision
tree21. This simple
tree design was chosen for its clarity. More sophisticated Markov modeling
techniques were not utilized because they require a level of detailed data not
available in the published literature pertaining to total ankle arthroplasty
and ankle fusion.
This model assumed a target population of patients seeking operative
treatment for ankle osteoarthritis at the age of fifty-five years. This age is
consistent with the typical age at which surgical intervention is required for
end-stage disease refractory to nonoperative
treatment1,13,19,20.
The time horizon of the model followed events through the remaining life
expectancy of twenty-five years for this age
group30.
The literature was reviewed to construct the event pathways following ankle
fusion. A table in the Appendix summarizes the data used to identify major
short and long-term complications following fusion. The primary short-term
complications included in the model were nonunion and below-the-knee
amputation. Reported nonunion rates have ranged from 0% to 41%, with most
studies demonstrating rates similar to the reference case assumption of
20%2,31-38.
The rate of amputation following fusion has been reported to be 0% to 15%,
with the rates in most reports close to the baseline assumption of 5% used in
this
study2,31-38.
Recent studies have highlighted the importance of adjacent joint arthritis as
a long-term complication of fusion. Some authors have reported hindfoot
arthritis to be an infrequent long-term complication, whereas others have
found it to be nearly
universal1,2.
The reference case in our study used an intermediate estimate of 50% for the
rate of adjacent joint arthritis. The uncertainty of this estimate was
addressed in the sensitivity analysis.
The event pathways following ankle replacement were also constructed
following the literature review. A second table in the Appendix summarizes the
data used to model outcomes following total ankle arthroplasty. Short-term
complications identified in these studies were categorized in the decision
model as requiring a reoperation, a revision to a fusion, or a below-the-knee
amputation. On the basis of published reports, an overall complication rate of
20% was used in the reference case. Several authors have reported similar
complication rates, with most falling in the range of 5% to
25%11,13,17,19,20.
The primary long-term complication of arthroplasty is eventual failure of
the implant, requiring revision. The long-term durability of newer designs of
ankle prostheses has not yet been reported and remains unknown. Several
studies have indicated high rates of prosthetic survival at five years
post-operatively11,13,19,20.
Detailed survival data over longer periods are not yet available. A long-term
duration of survival of ten years was extrapolated for use in the reference
case. This value was chosen to reflect the assumption that the new generation
of total ankle prostheses will prove to be more durable than previous
generations but probably will not survive for as long as total hip and knee
replacements39,40.
Given the uncertainty of this assumption, the effects on the
cost-effectiveness of total ankle arthroplasty of both shorter and longer
durations of survival were examined in the sensitivity analysis.
Effectiveness
This study is a special type of cost-effectiveness analysis termed
cost-utility analysis. Cost-utility analyses are differentiated by
the fact that effectiveness is measured in units that incorporate a subjective
measure of utility, such as quality-adjusted life years. The treatment of
ankle arthritis has minimal effects on patient survival but does result in
major changes in the quality of life of patients. The use of quality-adjusted
life years to measure effectiveness allows the survival of patients in
different health states to be corrected for health-related quality of life
with use of a utility
factor21.
Utility factors were assigned to all health states in the model to adjust
survival for quality of life. The reference case guidelines define utility
along a continuum, with a factor of 1.0 representing perfect health and a
factor of 0.0 representing
death21. Specific
utility values for each health state in this study were assigned on the basis
of a literature review. Table I
lists the health states included in the decision model along with their
corresponding utility values. Large-scale studies have used questionnaires to
establish utility values for a variety of health states. Arthritis has
consistently been shown to have a utility value near 0.7; similar values have
been reported for the utility of conditions resulting in impairment of a lower
extremity41-44.
On the basis of these studies, ankle arthritis was assigned a utility value of
0.7.
The actual long-term functional outcome and relative utility of total ankle
arthroplasty and ankle fusion have not been documented. Initial clinical
results have indicated that ankle replacement can result in a high level of
function11,13,14,19.
On the basis of these studies, we used a utility value of 0.9 for successful
total ankle replacement in this analysis. Ankle fusion was given a utility
value of 0.8, which is intermediate between the values for arthritis and a
successful arthroplasty. These values reflect the assumption in the reference
case that ankle replacement results in a better functional outcome than does
ankle fusion. This assumption is supported by cadaver studies that have shown
that ankle arthroplasty results in better kinematics and range of motion than
does ankle
fusion45,46.
However, these theoretical functional advantages of total ankle replacement
have not been definitively documented in long-term follow-up studies. The
resulting uncertainty of the utility values used for total ankle replacement
and ankle fusion was addressed in the sensitivity analysis.
Arthritis in the hindfoot following ankle fusion was given a utility value
of 0.7, equivalent to the baseline value for ankle
arthritis41-44.
Previously published data were used to assign a utility value of 0.6 to
below-the-knee amputation and 0.5 to periods of surgery and postoperative
recovery following
complications41-44.
Nonunion following ankle fusion was assigned a utility value (0.75) lower than
that of a successful ankle fusion. This is consistent with the limited data
published on the outcomes of ankle
nonunions31,34,35,37,38.
There were also limited data available with which to assign utility values to
revision arthroplasty and fusion after arthroplasty. These revision procedures
were assigned values (0.8 and 0.75, respectively) lower than those of the
corresponding primary operations. This practice is supported by published
reports of the clinical outcomes following revision of a failed ankle
arthroplasty47-49.
Costs
Gross-costing methodology was used to estimate the direct costs of both
total ankle arthroplasty and ankle fusion. This methodology relies on global
Medicare charge and reimbursement data to approximate the direct treatment
costs of various
procedures21.
Indirect costs such as lost productivity were not included in this analysis.
The surgical interventions included in the decision model were assigned their
appropriate ICD-9 (International Classification of Diseases—Ninth
Revision), DRG (Diagnosis-Related Groups), and CPT (Current Procedural
Terminology) codes. Gross cost estimates were then determined for acute care
hospitalizations and physician services on the basis of charge and
reimbursement data for these codes. The cost estimates for the DRG and CPT
codes are shown in Table
II.
Gross cost estimates for acute-care hospitalizations were determined from
mean hospital charges for the DRG associated with each intervention. Mean
hospital charges were based on data from the Centers for Medicare and Medicaid
Services (CMS) reported for
199850,51.
These hospital charges were then corrected to reflect actual costs with use of
a global cost-to-charge ratio of 0.47. This cost-to-charge ratio was derived
with use of data from the Medicare MEDPAR
database50,51.
The MEDPAR data source is released annually by Medicare and provides charge
and cost data for each DRG. The MEDPAR data for 1998 pertaining to all United
States hospitals were used in this study. The average cost was divided by the
average charge for all DRGs to determine the global cost-to-charge
ratio51.
The gross costs for physician services were determined from the mean
Medicare reimbursement for the CPT code associated with each surgical
intervention. The mean reimbursement reported by CMS in 1998 for the Los
Angeles, California, carrier was utilized. This global reimbursement includes
preoperative care, surgical fees, and ninety days of post-operative
care52. All costs
in this study are denominated in constant 1998 U.S. dollars.
Discounting
Cost-effectiveness analysis requires that all future costs and health
consequences be discounted and stated in terms of their present-day values.
Discounting is performed to correct for the fact that costs that are deferred
to the future are preferable to immediate expenditures. Costs and health
effects were discounted in the reference case at a constant rate of 3%
annually. Sensitivity analyses were conducted with discount rates of 0% and
5%21,22.
Sensitivity Analysis
Sensitivity analysis was conducted to test the uncertainty of the reference
case results. Cost-effectiveness analysis combines information from several
data sources in order to generate estimates of the probability of different
outcomes and assign values to their utility and costs. Uncertainty about the
true values of these underlying parameters results in uncertainty about the
cost-effectiveness ratios generated in the reference case. Sensitivity
analysis is used to determine the impact of varying the assumed values for key
variables on the conclusions generated by the cost-effectiveness analysis.
Initially, cost-effectiveness ratios were calculated with use of the
reference case assumptions for both ankle fusion and ankle replacement.
Sensitivity analysis was then performed with use of different assumptions for
the values of the underlying variables. Both univariate and multivariate
sensitivity analyses were conducted by altering the underlying variables both
in isolation and in
combination21.
Several key variables were selected for sensitivity analysis. These included
the duration of prosthetic function, the comparative quality of life following
ankle replacement and fusion, the rate of hindfoot arthritis following ankle
fusion, the probability of surgical complications, the assumed discount rate,
and the costs of the surgical procedures.
Reference Case Results
Outcomes
In the reference case, the pathway of total ankle replacement resulted in
14.23 quality-adjusted life years whereas that of ankle fusion resulted in
13.71 quality-adjusted life years. This led to an incremental gain of 0.52
quality-adjusted life years when arthroplasty was chosen over fusion. This
represents a 3.8% gain in efficacy.
Costs
Total ankle replacement generated total expected lifetime treatment costs
of $16,568. This exceeded the total lifetime treatment costs of $6990
following ankle fusion. As a result, ankle arthroplasty had a net incremental
increase in lifetime treatment costs of $9578.
Cost-Effectiveness Ratio
When chosen over ankle fusion, total ankle replacement had an incremental
cost-effectiveness ratio in the reference case of $18,419 for each
quality-adjusted life year gained. This ratio reflects the gain of 0.52
quality-adjusted life years at an incremental cost of $9578.
Sensitivity Analysis
Effect of the Functional Life of the Prosthesis
On the basis of published reports of promising results at five years
following total ankle arthroplasty, the reference case in this analysis
assumed a duration of survival of the prosthesis of ten
years11,14,19,20.
However, there remains substantial uncertainty about the durability of the
newer ankle prostheses; the actual durability may eventually prove to be
better or worse than this reference case assumption. Given this uncertainty, a
sensitivity analysis was conducted for the variable of duration of prosthetic
function. The analysis was repeated to determine the effect of shorter and
longer prosthetic survival periods on the cost-effectiveness gains derived
from total ankle replacement. Figure
2 illustrates the relationship of the assumed duration of
prosthetic function to the incremental cost-effectiveness ratio of ankle
replacement.
Sensitivity analysis with the assumption of a longer prosthetic survival
time required modifying the event pathway following successful arthroplasty.
The reference case assumed that patients would undergo revision to fusion
following failure of a primary arthroplasty. Longer prosthetic survival makes
revision arthroplasty a more realistic option. The decision tree probabilities
were changed to create a situation in which patients became candidates for
revision arthroplasty when the primary prosthesis survived longer than the
reference case value of ten years. The cost-effectiveness of ankle
arthroplasty improved when the prosthesis was assumed to function for longer
periods. The incremental cost-effectiveness ratio of total ankle arthroplasty
fell below $8500 per quality-adjusted life year gained when prosthetic
survival was assumed to approach fifteen years or longer.
Conversely, a shorter functional life of the prosthesis decreased the
cost-effectiveness of total ankle replacement when compared with ankle fusion.
The cost-effectiveness ratio rose above $25,000 per life year gained when
prosthetic durability fell below nine years. The threshold of $50,000 per life
year gained was passed as prosthetic functional life was assumed to approach
seven years. There was no incremental gain in health-related quality of life
when prosthetic survival was assumed to be less than 5.5 years.
Effect of Functional Outcome Following Ankle Replacement
Multivariate sensitivity analysis was conducted with use of a range of
values for the utility of both total ankle replacement and ankle fusion. The
reference case of this analysis assigned a higher utility to ankle replacement
than to ankle fusion. This utility difference reflects the assumption that
gait kinematics and functional results following ankle replacement are better
than those following ankle
fusion11,20,45,46.
However, these theoretical functional advantages of total ankle arthroplasty
have not been definitively documented in published trials. The actual
utilities remain uncertain given the limited long-term data available.
Sensitivity analysis was used to examine the effect of altering the assumed
utilities of total ankle replacement and fusion on the incremental
cost-effectiveness of ankle arthroplasty.
The incremental gain in quality-adjusted life years from total ankle
replacement became negligible when the utility of arthroplasty was assumed to
be similar to that of fusion. Conversely, the cost-effectiveness of
arthroplasty improved when it was assumed to result in a larger improvement in
utility than that following ankle fusion.
Figure 3 graphs the effect of
altering the utility of ankle arthroplasty on its incremental
cost-effectiveness ratio.
Effect of Rate of Symptomatic Hindfoot Arthritis
There is substantial variation in the reported rate of symptomatic hindfoot
arthritis following ankle fusion. The reference case analysis assumed a
midrange rate of 50%. Sensitivity analysis was conducted with use of a range
of values for the rate of hindfoot arthritis as a late complication of fusion.
Figure 4 illustrates the effect
of altering the assumed rate of adjacent joint arthritis on the incremental
cost-effectiveness ratio of total ankle arthroplasty. The cost-effectiveness
of the arthroplasty improved as the rate of hindfoot arthritis following ankle
fusion was assumed to be higher and declined when the rate was assumed to be
lower than the reference case value.
Effect of Complication Rate of Total Ankle Arthroplasty
The complication rate following total ankle replacement has not been firmly
established. The reference case assumed a complication rate of 20%; most
published studies have demonstrated a rate of between 5% and
25%11,13,17,19,20.
Sensitivity analysis using a wide range of values for the assumed complication
rate did not result in large variations in the cost-effectiveness ratio of
total ankle arthroplasty. The cost-effectiveness ratio remained close to
$13,000 per quality-adjusted life year when the complication rate was assumed
to be as low as 5%. The cost-effectiveness ratio did not rise above $25,000
per quality-adjusted life year gained at an assumed complication rate of
25%.
Effect of Discount Rate
Analyses using discount rates of 0% and 5% in addition to the baseline
assumption of 3% were conducted. These variations in the discount rate did not
have a large impact on the cost-effectiveness ratio. A discount rate of 0% led
to an incremental cost-effectiveness ratio for total ankle arthroplasty of
$21,561 per quality-adjusted life year gained. A discount rate of 5% resulted
in a ratio of $17,252 per quality-adjusted life year gained.
Effect of Altering Assumptions of Costs
Sensitivity analysis was used to determine the effect of inaccuracies in
cost assumptions on the incremental cost-effectiveness ratio of total ankle
arthroplasty. Gross-costing methodology was chosen in this study to provide
estimates of the typical costs of care for the procedures included in the
decision model. This methodology may provide less precise estimates of costs
than would more detailed micro-costing
techniques21. This
uncertainty was addressed by calculating the cost-effectiveness ratio of total
ankle arthroplasty over a wide range of values for the underlying cost
assumptions.
Sensitivity analysis demonstrated that altering cost estimates over a large
range did not change the qualitative results of this study. Multivariate
analyses were performed to examine the effect of proportional changes in the
assumed value of all cost estimates. Simultaneously increasing all cost
estimates by up to 100% of their reference case values did not cause the
incremental cost-effectiveness ratio of total ankle arthroplasty to rise above
$40,000 per quality-adjusted life year. Similar results were seen in
univariate analysis with use of a range of values for the cost of ankle
replacement (Fig. 5). The
incremental cost-effectiveness ratio remained below $40,000 per
quality-adjusted life year when the cost assumption for total ankle
arthroplasty was increased by up to 100% of the reference case value.
This study demonstrates that the currently available literature has not yet
established that total ankle arthroplasty predictably results in levels of
durability and function that demonstrate it to be cost-effective at this time.
Sensitivity analysis identified a threshold of seven years for prosthetic
survival before the incremental cost-effectiveness ratio of total ankle
arthroplasty begins to fall below a value of $50,000 per quality-adjusted life
year gained. This ratio falls below $25,000 per quality-adjusted life year
gained when prosthetic survival is assumed to be nine years or longer.
Specific thresholds for determining favorable cost-effectiveness ratios have
not been established. However, several authors have suggested that
cost-effectiveness ratios below $50,000 per quality-adjusted life year gained
are indicative of moderately cost-effective interventions. Interventions with
an incremental cost-effectiveness ratio of $25,000 or lower per
quality-adjusted life year gained are generally considered to be highly
cost-effective21.
While several studies have shown promising five-year survival rates for total
ankle replacements, the published data have not yet adequately documented
survival over longer
periods1,11,13,14,19,20.
The use of total ankle arthroplasty will be better justified if it is shown to
reliably result in prosthetic survival beyond the range of seven to nine
years.
This analysis also highlights the importance of the relative functional
outcome of ankle arthroplasty and ankle fusion in determining the
cost-effectiveness of ankle replacement. Total ankle arthroplasty has been
shown in cadaver studies to have potential advantages with regard to function
when compared with ankle fusion, and initial clinical studies have shown
promising functional results of total ankle
arthroplasty11,13,14,19,20,45,46.
These findings are reflected in the reference case assumption that total ankle
arthroplasty results in the higher utility value of 0.9 while ankle fusion is
assigned a value of 0.8. However, the long-term utilities of ankle
arthroplasty and ankle fusion have not been documented and remain unknown.
Sensitivity analysis was used to examine the effect of the uncertainty of
these utility values on the results of this study. When more modest
assumptions are used for the gain in utility following total ankle
replacement, the incremental cost-effectiveness ratio of arthroplasty rises
and arthroplasty becomes less cost-effective. The threshold ratio of $25,000
per quality-adjusted life year is crossed when the utility of ankle
arthroplasty falls below 0.88. The cost of ankle arthroplasty exceeds $50,000
per quality-adjusted life year when its utility is less than 0.86. The
literature has not yet definitively shown that total ankle arthroplasty
results in long-term functional outcomes that justify assigning it a utility
value higher than that of ankle fusion. The use of total ankle arthroplasty
will be more justified if its theoretical functional advantages are supported
by long-term clinical trials that provide more precise estimates of the
utility values of total ankle arthroplasty and ankle fusion.
The reference case in this study did demonstrate the potential for total
ankle replacement to be a cost-effective alternative to ankle fusion. The
reference case results depend on an extrapolation of prosthetic survival of
ten years and on the assumption that total ankle arthroplasty results in
better function than does ankle fusion. With use of these reference case
assumptions, ankle replacement resulted in an incremental cost-effectiveness
ratio of $18,419 per quality-adjusted life year gained when chosen over ankle
fusion. This cost-effectiveness ratio compares favorably with that of other
medical and surgical interventions. An examination of total hip replacement
demonstrated a cost-effectiveness ratio of $4600 (1991 U.S. dollars) per
quality-adjusted life
year53. The authors
of that study concluded that this ratio supported hip replacement as a highly
cost-effective procedure. They cited examples of other medical interventions
that are widely accepted as cost-effective, including coronary artery bypass
grafting ($8100 per quality-adjusted life year) and medical treatment of
hypertension ($24,900 per quality-adjusted life year). The authors of a
separate study advocated lumbar discectomy as a cost-effective procedure, with
a ratio of $29,200 (1993 U.S. dollars) per quality-adjusted life year
gained54. The
reference case in the present study demonstrates the potential of total ankle
replacement to have a similar level of cost-effectiveness if long-term
clinical results meet or exceed the reference case assumptions.
One potential source of error in this analysis is inaccuracy in the
construction of the decision model. A simple branching design was used to
model the outcomes following total ankle arthroplasty and ankle fusion. More
sophisticated Markov modeling techniques were not utilized because they
require a level of detailed data not available in the published literature
pertaining to ankle arthroplasty and ankle fusion. The use of alternative
techniques such as Markov modeling may increase fidelity in representing
probabilities and utilities as they change over time. Despite its potential
advantages, the inclusion of Markov processes has not been shown to change the
qualitative results of cost-effectiveness
analyses21,29.
Another potential source of uncertainty in this analysis arises from the
estimation of the costs, effectiveness, and utilities of health states
included in the decision model. The study identified key variables that are
unknown, including the durability of total ankle prostheses as well as the
relative long-term utility of ankle fusion and ankle replacement. Threshold
values for these unknown variables that would indicate the cost-effectiveness
of total ankle arthroplasty were identified in sensitivity analysis. There is
also some level of uncertainty about several of the other utility values used
in this study, including those for ankle arthritis, hindfoot arthritis, and
nonunion. These values were extrapolated from studies that determined the
utility of arthritis and lower-extremity
impairment41-44.
This analysis would benefit from empirical studies that more directly measure
the long-term utility of these conditions. Similarly, the gross-costing
methods used in this study may not provide as precise an estimate of costs as
would more detailed micro-costing techniques. The effect of the uncertainty
about these variables was examined with use of sensitivity analysis.
Sensitivity analyses using a range of values for these utility and cost
estimates did not change the qualitative results of this study.
Total ankle arthroplasty represents a potential technological advance in
the treatment of ankle arthritis. Despite a lack of long-term data, the
interest in implementing this new technology is growing. The dissemination of
total ankle arthroplasty will be more justified if long-term clinical trials
document levels of prosthetic survival and function that meet or exceed the
thresholds identified in this study. These thresholds can help orthopaedic
surgeons to evaluate the results of clinical trials as they become available
and help them to decide whether to perform total ankle arthroplasty.
Tables summarizing data from the literature on outcomes of ankle fusion and
total ankle arthroplasty are available in the electronic versions of this
article, on our web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?