Abstract
Background: The efficacy of total ankle replacement compared with
that of ankle fusion continues to be one of the most debated topics in foot
and ankle surgery. The purpose of this study was to determine whether there
are sufficient objective cumulative data in the literature to compare the two
procedures. A systematic review of the literature addressing the intermediate
and long-term outcomes of interest in total ankle arthroplasty and ankle
arthrodesis was performed.
Methods: A comprehensive search of MEDLINE for all relevant articles
published in English from January 1, 1990, to March 2005 was conducted.
Additionally, relevant abstracts from the 2003 and 2004 annual proceedings of
major orthopaedic meetings were eligible. Two reviewers evaluated each study
to determine whether it was eligible for inclusion and collected the data of
interest. Meta-analytic pooling of group results across studies was performed
for the two procedures. The analysis of the outcomes focused on
second-generation ankle implants.
Results: The systematic review identified forty-nine primary
studies, ten of which evaluated total ankle arthroplasty in a total of 852
patients and thirty-nine of which evaluated ankle arthrodesis in a total of
1262 patients. The mean AOFAS (American Orthopaedic Foot and Ankle Society)
Ankle-Hindfoot Scale score was 78.2 points (95% confidence interval, 71.9 to
84.5) for the patients treated with total ankle arthroplasty and 75.6 points
(95% confidence interval, 71.6 to 79.6) for those treated with arthrodesis.
Meta-analytic mean results showed 38% of the patients treated with total ankle
arthroplasty had an excellent result, 30.5% had a good result, 5.5% had a fair
result, and 24% had a poor result. In the arthrodesis group, the corresponding
values were 31%, 37%, 13%, and 13%. The five-year implant survival rate was
78% (95% confidence interval, 69.0% to 87.6%) and the ten-year survival rate
was 77% (95% confidence interval, 63.3% to 90.8%). The revision rate following
total ankle arthroplasty was 7% (95% confidence interval, 3.5% to 10.9%) with
the primary reason for the revisions being loosening and/or subsidence (28%).
The revision rate following ankle arthrodesis was 9% (95% confidence interval,
5.5% to 11.6%), with the main reason for the revisions being nonunion (65%).
One percent of the patients who had undergone total ankle arthroplasty
required a below-the-knee amputation compared with 5% in the ankle arthrodesis
group.
Conclusions: On the basis of these findings, the intermediate
outcome of total ankle arthroplasty appears to be similar to that of ankle
arthrodesis; however, data were sparse. Comparative studies are needed to
strengthen this conclusion.
Level of Evidence: Therapeutic Level IV. See Instructions
to Authors for a complete description of levels of evidence.
The development of total ankle implants began in the 1970s as an
alternative to ankle arthrodesis. First-generation cemented total ankle
implants, constrained and unconstrained, often failed, primarily as a result
of poor implant design, loosening, and
instability1,2.
Despite these results, the search for a successful ankle prosthesis continued,
leading to the development of a second generation of total ankle implants with
numerous modifications, including a semiconstrained cementless design, and
mobile and fixed-bearing
designs1,3,4.
The relative benefits of total ankle arthroplasty and ankle fusion continue
to be one of the most debated topics in foot and ankle surgery. Proponents for
each procedure advocate their point of view very strongly, but most of the
information is anecdotal and based on personal experience. The purpose of this
study was to determine whether there are sufficient objective cumulative data
in the literature to compare the two procedures.
We performed a systematic review of the literature addressing the
intermediate and long-term outcomes of total ankle arthroplasty and ankle
arthrodesis in terms of ankle function, pain, revision, conversion to
arthrodesis, implant survival, and quality of life.
Standard systematic review methods were
used5,6.
Initially, a prospective protocol was written to describe the objectives,
search criteria, study selection criteria, data elements of interest, and
plans for analysis. According to the protocol, a broad search of the
English-language literature spanning fifteen years, from January 1, 1990, to
March 24, 2005, was conducted. MEDLINE was searched through PubMed with use of
the following search strategies: (1) "ankle replacement" OR
"ankle prosthesis" or "ankle fusion"; (2)
(arthroplasty, replacement [MeSH] AND "ankle") OR (Arthrodesis
[MeSH] AND "ankle"); (3) 1 OR 2.
A manual reference check of all accepted papers and recent reviews was
performed to supplement the electronic searches and to identify any additional
potentially relevant studies. In addition to the published literature,
abstracts from the 2003 and 2004 annual proceedings of the American Academy of
Orthopaedic Surgeons (AAOS) and American Orthopaedic Foot and Ankle Society
(AOFAS) meetings were searched for eligible studies. Eligibility requirements
were the same for abstracts and full publications.
For a study to satisfy the criteria for inclusion, the authors had to have
reported on (1) total ankle arthroplasty or tibiotalar ankle arthrodesis
surgery, (2) outcomes of interest at two years or more after the operation,
and (3) at least ten patients in the treatment group. Studies in which all
patients had been treated with only revision total ankle arthroplasty or
revision arthrodesis were excluded. There were two reviewers, who had to agree
on all accepted and rejected studies. All study designs were eligible,
including randomized controlled trials, prospective and retrospective
nonrandomized controlled trials, and uncontrolled case series. Multiple
publications on the same patient population were pooled as one study (kinship)
to the extent possible to avoid double-counting of patients. Protocol-defined
data elements from each eligible study were extracted by one researcher onto a
data-extraction form and were confirmed by a second researcher, who checked
all extracted data against the original report. Differences were resolved
prior to data entry.
Statistical Methods
Outcomes of interest included patient-reported outcome scores, mobility,
function, pain, revision or conversion to an arthrodesis, implant survival,
and prosthetic loosening. In addition, the overall postoperative outcome,
based on the patient's assessment, was recorded. Study, patient, and treatment
characteristics were summarized with use of basic descriptive statistics. The
number of patients enrolled (or randomized) was used as a denominator in the
analysis of baseline patient characteristics in the assessment of
patient-reported efficacy outcomes. The number of ankles evaluated was used as
the denominator in the analyses of all other efficacy outcomes. Follow-up
scores on the AOFAS Ankle-Hindfoot Scale and on other scales evaluating ankle
pain, function, alignment, and range of motion were summarized for both total
ankle arthroplasty and ankle arthrodesis. Similarly, overall patient outcomes
were summarized with use of four categories: excellent, good, fair, and poor.
The definitions of these four categories were for the most part uniform across
studies. The AOFAS Scale (with excellent defined as 90 to 100 points; good, as
75 to 89 points; fair, as 50 to 74 points; and poor, as <50 points) was the
most commonly used rating system in the studies of total ankle arthroplasties.
The Mazur Score (with excellent defined as 80 to 90 points; good, as 70 to 79
points; fair, as 60 to 69 points; and poor, as <60 points) was most
commonly used in the studies of arthrodeses. In order to include all available
data, overall outcomes were also categorized as good (including excellent and
good results) and poor (including fair and poor results).
Data were summarized in two ways. First, raw means and counts were weighted
by sample size and calculated for all outcomes. Second, selected efficacy
outcomes were synthesized with random-effects meta-analytic pooling of group
results across
studies7,8.
In random-effects meta-analyses, study results are weighted by a combination
of study size and between-study variation. When there was no variation across
study results that would not be expected due to random chance, the weighted
results were equivalent to those found by weighting by study sample size. The
presence of variation across study results suggests that important study-level
factors were impacting results, causing large studies to be given somewhat
less weight and, concomitantly, smaller studies to be given somewhat more
weight. We report the efficacy results associated with second-generation
implants, including the Agility and New Jersey LCS (DePuy, Warsaw Indiana),
Buechel-Pappas (Endotec, South Orange, New Jersey), TNK (Kyocera, Kyoto,
Japan), STAR (Waldemar Link, Hamburg, Germany), and Salto (Tornier,
Montbonnot, France) prostheses. Similar efficacy outcomes were analyzed, with
use of the same meta-analytic techniques, in the review of the ankle
arthrodesis studies. No formal comparison of efficacy or safety outcomes
between total ankle arthroplasty and ankle arthrodesis was performed.
All calculations were performed with use of SAS software (version 8.1;
Cary, North Carolina) and SPSS software (version 14.0; Chicago, Illinois).
Studies
The initial search yielded 460 citations, of which 167 were retrieved for
further review. Sixty-two full publications and five meeting abstracts met all
eligibility criteria for inclusion into the database (see Appendix). Eighteen
of these were identified as kinship studies. The final database included
forty-nine primary studies spanning the years 1990 to 2005. The studies
included fifty-six treatment arms and 2114 patients. Study attrition is shown
in Figure 1. The
characteristics of the accepted studies are further detailed in
Table I.
Of the forty-nine studies, ten (20%) focused on total ankle arthroplasty
and thirty-nine (80%) described ankle arthrodesis; 852 patients underwent
total ankle arthroplasty and 1262 underwent arthrodesis. The majority of the
arthrodesis studies (56%) were published between 1990 and 1997, whereas all
ten of the total ankle arthroplasty studies were published more recently,
between 1998 and 2005. The vast majority of studies were from single centers
in the United States and Europe. Other geographic locations included Japan and
Taiwan.
There were no studies directly comparing total ankle arthroplasty and
arthrodesis. Three non-randomized controlled trials of ankle arthrodesis
compared different surgical techniques for arthrodesis. The duration of
follow-up ranged from two to nine years in the studies of total ankle
arthroplasty and from two to twenty-three years in the studies of arthrodesis.
The average follow-up time following the total ankle arthroplasties and the
arthrodeses was approximately five years.
All ten studies of total ankle arthroplasty (including a total of 852
patients) focused on second-generation total ankle implants. Information on
the implants and the methods of fixation used in the total ankle
arthroplasties is also displayed in Table
I.
Patients
Patient characteristics were minimally described in the studies. In those
in which the age of the patient population was reported, the mean overall age
was fifty-three years (mean range, eighteen to sixty-four years) (see
Appendix). The patients treated with total ankle arthroplasty were older (mean
age, fifty-eight years; mean range, forty-six to sixty-four years) than those
treated with ankle arthrodesis (mean age, fifty years; mean range, eighteen to
sixty-three years) and were predominately female (59%). The majority of the
patients in the arthrodesis studies were male (52%). Rheumatoid arthritis was
the primary indication for the total ankle arthroplasties (39%), whereas
posttraumatic arthritis was the primary indication for the arthrodeses
(57%).
Efficacy Outcomes
The reporting of efficacy outcomes was highly variable in both the total
ankle arthroplasty and the ankle arthrodesis studies; therefore, the analyses
performed in the present study were limited. The AOFAS Ankle-Hindfoot Scale
total score and pain, function, and alignment subscores; the Kofoed scores for
pain and function; and the range of motion (overall and plantar
flexion-dorsiflexion) were among the most frequently reported evaluation
methodologies. Table II
presents both the raw means for these outcomes and the pooled meta-analytic
results for the AOFAS total score and the range-of-motion outcomes for both
total ankle arthroplasty and ankle arthrodesis. The meta-analysis mean AOFAS
score was 78.2 points (95% confidence interval, 71.9 to 84.5) for the group
treated with total ankle arthroplasty and 75.6 points (95% confidence
interval, 71.6 to 79.6) for the group treated with ankle arthrodesis. The raw
mean scores on the individual AOFAS subscales for pain, function, and
alignment were 34.5, 37.4, and 9.4 points, respectively, in the group treated
with total ankle arthroplasty. In the arthrodesis group, the raw mean Kofoed
scores for pain and function were 32.5 and 15.7 points, respectively. Because
of the lack of reporting of variances, meta-analysis of the subscale scores
was not possible.
In the studies of total ankle arthroplasty in which excellent, good, fair,
and poor categories were used as overall outcome measures, the result was
excellent in 38% of the patients (95% confidence interval, 0% to 96.8%), good
in 30.5% (95% confidence interval, 21.0% to 39.9%), fair in 5.5% (95%
confidence interval, 0% to 16.9%), and poor in 24% (95% confidence interval,
0% to 72.9%). In the arthrodesis group, the pooled outcomes were excellent in
31% (95% confidence interval, 19.8% to 41.5%), good in 37% (95% confidence
interval, 26.4% to 47.3%), fair in 13% (95% confidence interval, 6.2% to
20.3%), and poor in 13% (95% confidence interval, 7.6% to 18.7%).
The authors of several studies used different categorizations of overall
patient outcomes (e.g., good-to-excellent results); therefore, in order to
include these studies in the analysis, outcomes were combined into two wider
categories: good (including both excellent and good results) and poor
(including both fair and poor results). The meta-analysis of this binary
outcome (Table III) showed
that, in the studies of total ankle arthroplasty, 78% (95% confidence
interval, 61.9% to 95.0%) of the patients had a good result and 22% (95%
confidence interval, 4.9% to 38.6%) had a poor result. In the arthrodesis
studies, 73% (95% confidence interval, 61.2% to 84.1%) had a good result and
27% (95% confidence interval, 16.0% to 38.8%) had a poor result.
Data on implant survival, revisions, conversions to arthrodesis, and other
clinical outcomes after the total ankle arthroplasties and ankle arthrodeses
are presented in the Appendix. The five-year and ten-year implant survival
rates following total ankle arthroplasty were 78% (95% confidence interval,
69.0% to 87.6%) and 77% (95% confidence interval, 63.3% to 90.8%),
respectively. A revision during the follow-up period was required in 7% (95%
confidence interval, 3.5% to 10.9%) of the patients who had undergone a total
ankle arthroplasty. The most common reason for revision was loosening and/or
subsidence (28%). Five percent (95% confidence interval, 2.0% to 7.8%) of the
total ankle arthroplasties were converted to arthrodeses, with the main reason
for conversion being loosening and/or subsidence (50% of all conversions).
Below-the-knee amputation was performed in 1% of the patients treated with
total ankle arthroplasty.
Nonunion was observed in 10% (95% confidence interval, 7.4% to 12.1%) of
the patients treated with ankle arthrodesis. Nine percent (95% confidence
interval, 5.5% to 11.6%) of the arthrodesis group underwent revision,
primarily because of nonunion (the indication for 65% of all revisions of
arthrodeses). Five percent of the patients treated with ankle arthrodesis
underwent a below-the-knee amputation. Significant heterogeneity was detected
in almost all of these meta-analyses; therefore, the results must be
interpreted with caution.
To our knowledge, this is the first systematic review of the intermediate
and long-term outcomes of total ankle arthroplasty and ankle arthrodesis.
Baseline differences in the patient populations and the small number of
studies contributing to each analysis do not permit formal comparison between
the two surgical procedures. However, the intermediate and long-term outcomes
analyzed in this review do suggest that total ankle arthroplasty is comparable
with ankle arthrodesis.
The strengths of this review include the clear definition of the research
question to eliminate bias in the selection of the studies, adherence to an
explicit research protocol that was developed prior to the analysis, the
comprehensive nature of the literature search (with use of both electronic
databases and manual bibliography searches), consensus between the two
reviewers with regard to all data elements prior to entry into the database,
and a quality-control review of all results. However, despite the strengths of
the review process, they cannot overcome the inherent weaknesses in the
literature.
The primary limitation of this review is that a direct comparative
meta-analysis of total ankle arthroplasty and arthrodesis was not possible
because there were no head-to-head trials. We were only able to perform a
pooled meta-analysis across all studies, with many studies being devoid of key
data elements, including methodology reporting and baseline patient
information. In addition, the variability of the reporting of outcomes of
interest limited the number of studies for each meta-analysis. In fact,
differences in patient populations, variability of surgical procedures, and
differences in outcome evaluation tools and study follow-up times may all be
partially responsible for heterogeneity among these studies. In future
studies, it will be important to use uniform evaluation tools to make
comparisons easier.
This study exposed the major lack of objective, prospective, and controlled
data on either procedure. The findings of this review demonstrate that the
available data on the results of these procedures are based predominantly on
retrospective uncontrolled case series from single institutions. The sample
sizes in many of the studies were small. Therefore, prospective studies are
needed to compare the two procedures in similar patient groups.
Despite these limitations, this study provides evidence that both
procedures yield satisfactory results, and the data suggest equivalence
between the procedures, negating the poor connotations associated with ankle
arthroplasty due to failures associated with first-generation implants. In
fact, while all patients may not be candidates for ankle arthroplasty, it
should be considered as a treatment option for those with ankle arthritis,
with the clinician being allowed the latitude to determine appropriate
indications.
Tables presenting the patient characteristics and the clinical outcomes
from the combined studies as well as a list of the studies that were evaluated
are available with 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). ?
Easley ME, Vertullo CJ, Urban WC, Nunley
JA. Total ankle arthroplasty. J Am Acad Orthop Surg.2002;10:
157-67.10157
2002
[PubMed]
Neufeld SK, Lee TH. Total ankle
arthroplasty: indications, results, and biomechanical rationale. Am J
Orthop.2000;29:
593-602.29593
2000
Cook RA, O'Malley MJ. Total ankle
arthroplasty. Orthop Nurs.2001;20:
30-7.2030
2001
[PubMed][CrossRef]
Pyevich MT, Saltzman CL, Callaghan JJ,
Alvine FG. Total ankle arthroplasty: a unique design. Two to twelve-year
follow-up. J Bone Joint Surg Am.1998;80:
1410-20.801410
1998
[PubMed]
Cook DJ, Mulrow CD, Haynes RB.
Systematic reviews: synthesis of best evidence for clinical decisions.
Ann Intern Med.1997;126:
376-80.126376
1997
[PubMed]
Alderson P, Green S, Higgins JPT,
editors. Cochrane Reviewers' Handbook 4.2.1 [updated December 2003]. In:
The Cochrane Library, Issue 1. Chichester, UK: John Wiley and
Sons; 2004.
2004
DerSimonian R, Laird N. Meta-analysis in
clinical trials. Control Clin Trials.1986;7:
177-88.7177
1986
[PubMed][CrossRef]
Hedges LV, Olkin I. Statistical
methods for meta-analysis. Orlando, FL: Academic Press;
1985. p 230-57.230
1985