The Norwegian Orthopaedic Association started a national register for total
hip replacement in
19871. In January
1994, the register was expanded to include all artificial joint replacements,
including those of the
knee2,3.
One of the aims of the register is to detect inferior implants, cements, and
techniques as early as possible.
Unicompartmental knee replacements were popular in the 1970s and 1980s in
Europe, but, because of problems with fixation and the high number of
failures, the use of these implants was reduced in the
1990s4. In the late
1990s, good ten-year results of unicompartmental knee replacement were
reported from single centers in both the United States and
Britain5,6.
These results, together with new instrumentation for minimally invasive
surgery, have renewed interest in this procedure, but there have been numerous
reports of concern about the durability of this type of
prosthesis7-10.
The aim of the present study was to compare early failure rates and failure
mechanisms of primary cemented unicompartmental knee replacements with those
of primary cemented tricompartmental total knee replacements with use of the
nationwide prospective observational knee implant register of all Norwegian
hospitals.
After each operation, a standard form is filled out by the surgeon and is
sent to the
register11. The
reporting is similar to that for hip
replacement1.
Stickers with catalogue numbers are delivered by the manufacturers along with
the implants and are attached to the form by the operating surgeon. Femoral,
tibial baseplate, tibial polyethylene insert, and patellar components are
registered separately.
Information on revisions, defined as a surgical removal or exchange of a
part of the implant, or of the whole implant, was linked to data on the
primary operation with use of the unique identification number assigned to
each inhabitant of Norway.
The types of primary cemented unicompartmental knee replacements used in
Norway were the MOD III (Smith and Nephew, Memphis, Tennessee), Genesis
Uni (Smith and Nephew), Oxford II and III (Biomet, Bridgend, South
Wales, United Kingdom), Duracon all-polyethylene tibial Uni (Stryker,
Berkshire, United Kingdom), Miller-Galante all-polyethylene tibial Uni
(Zimmer, Warsaw, Indiana), and Preservation (DePuy, Leeds, United Kingdom)
prostheses as well as other prostheses. There were no uncemented
unicompartmental knee replacements. The size of the Preservation knee group
and the "other" group were too small for separate analyses.
The survival of cemented unicompartmental knee replacements and cemented
patellar resurfaced total knee replacements, inserted in the period from
January 1, 1994, to December 31, 2004, was compared at five, seven, and ten
years of follow-up. For primary unicompartmental knee replacements, we
compared the time until revision for each brand of prosthesis used. The
five-year survival rates for the different total knee replacement brands used
in Norway were reported in an earlier
study11. The
different causes of failure leading to revision were compared for cemented
unicompartmental knee replacements and cemented total knee replacements. The
surgeon could report one or more causes of failure leading to revision.
Possible causes were aseptic loosening of the femoral, tibial, or patellar
component; dislocation or instability; malalignment; deep infection;
periprosthetic fracture; pain; wear of a tibial insert; or other causes. To be
classified as having a revision because of pain alone, no other reason for
revision could be marked. When seen in combination with any other cause,
infection was considered as the primary reason for revision.
The reports from the Norwegian Arthroplasty Register were compared with the
compulsory national hospital administrative database, the Norwegian Patient
Register. An estimated 99% of the primary and 97% of the revision knee
prostheses were reported to the Norwegian Arthroplasty Register as compared
with the Norwegian Patient Register during the years 1999 through
200212.
Statistical Analysis
Prosthesis survival was calculated with use of the Kaplan-Meier method.
Because of the low number of prostheses at risk after eleven years of
follow-up, survival results were estimated at ten years, and for two of the
prostheses with the shortest follow-up (the Oxford III and the Miller-Galante
prostheses), the results were estimated at five years. The median duration of
follow-up was calculated with use of the reverse Kaplan-Meier
method13. The
survival curves were stopped when the number of knees at risk was lower than
twenty. Patients who died or emigrated during the follow-up period were
identified from files provided by Statistics Norway, and the follow-up time
for prostheses in these patients was censored at the date of death or
emigration. A Cox multiple regression model was used to study the relative
risks of revision among unicompartmental knee replacement prosthesis brands,
to study differences between unicompartmental knee replacements and total knee
replacements, and to adjust for potential confounding by age (sixty years old
or less, sixty-one to sixty-nine years old, and seventy years old or more),
gender, and diagnosis (primary gonarthrosis, rheumatoid arthritis, sequela
after fracture, sequela after ligamentous instability, sequela after meniscal
injury, and others). The confounding effect of age was further investigated as
a continuous variable within age-categories. The average annual volume of
surgery at individual hospitals (zero to nine operations, ten to nineteen
operations, and twenty to forty-nine operations) was tested in a Cox model
with adjustment for age, gender, and prosthesis design.
The statistical analyses were performed with use of SPSS software (Advanced
Statistics 13.0; SPSS, Chicago, Illinois) and S-PLUS 2000 (Insightful,
Seattle, Washington). Two-sided p values of <0.05 were considered
significant.
Epidemiology of Knee Replacement Surgery
During the study period, 19,669 primary knee replacements were reported; of
these, 3032 were cemented total knee replacements with patellar resurfacing
and 2288 were unicompartmental knee replacements
(Table I). The majority of the
rest of the replacements were cemented total knee replacements without
patellar resurfacing. Unicompartmental knee replacements accounted for 15% of
the primary knee replacements in 2004 and 12% of the primary knee replacements
during the entire eleven-year study period. There was a decrease in use from
1994 to 1998, with the lowest annual percentage of 5% in 1997, and then an
increase in use. Ninety-eight percent of the total knee replacements were
posterior cruciate-retaining
designs11.
Patients managed with unicompartmental knee replacement were younger and
were more likely to be male in comparison with those managed with total knee
arthroplasty (Table I). In
patients managed with unicompartmental knee replacement, 89% had primary
osteoarthritis, 7.6% had sequelae after meniscal injury, 1.7% had sequelae of
a fracture, 1.1% had sequelae of osteochondritis, 1.4% had sequelae of
osteonecrosis, 0.5% had other diseases, and only 0.3% had rheumatoid
arthritis.
A higher proportion of the Oxford III and Miller-Galante prostheses were
implanted in patients who were sixty years old or less than was the case for
prostheses that had longer follow-up. An intact anterior cruciate ligament
after the operation was reported following 96% of the unicompartmental knee
replacements (see Appendix). All of the unicompartmental knee replacements
were cemented, with 89% being cemented with use of Palacos cement with
gentamicin (Schering-Plough, Kenilworth, New Jersey). Only three types of
prostheses were meniscal-bearing: the Oxford II, the Oxford III, and the
Preservation. The Duracon and the Miller-Galante unicompartmental knee
replacements were always used with an all-polyethylene tibial component, and
the Genesis and Preservation replacements were used with an all-polyethylene
tibial component in approximately one-quarter of the cases. The most common
polyethylene thicknesses for the tibial insert were 8 and 10 mm, except for
the Oxford II and III prostheses, for which 3.5, 4.5, and 5.5-mm-thick
polyethylene inserts were most common. The thinnest polyethylene components
(7.5 mm) for the MOD III implant were used at the beginning of the period. It
was not possible to determine whether the prosthetic components were placed in
the medial or the lateral compartment in two of the prosthesis designs at the
beginning of the registration period because these two implants could be used
both in the lateral and the medial compartment. The form was changed in 2001
to include registration of the lateral or medial compartment.
Volume of Surgery
The 2288 unicompartmental knee replacements had been performed in fifty-one
hospitals over an eleven-year period (see Appendix). On the average, there had
been four unicompartmental knee replacement operations per hospital per year
during the study period, and, at forty-three of the fifty-one hospitals, there
had been an average of fewer than ten procedures per year. Seven types of
unicompartmental knee replacements had been used in more than twenty knees
each. Four prostheses had been used in >100 knees. When we investigated all
prostheses and all hospitals, we found that the hospitals at which an average
of twenty to forty-nine knee operations had been performed per year (n = 3)
had a 40% lower revision risk as compared with those at which zero to nine
operations had been performed per year (n = 43) (relative risk, 0.6 [95%
confidence interval, 0.40 to 1.0]; p = 0.05). For the Miller-Galante knee
replacement, there was no hospital at which an average of more than ten
operations had been performed per year, and therefore the influence of
hospital surgery volume could not be tested for this implant. For the Oxford
III knee replacement, there was no difference, with the numbers available,
between hospitals at which zero to nine (n = 30), ten to nineteen (n = 4), or
twenty to forty-nine (n = 3) procedures had been performed per year.
Survival Rates
The ten-year survival rate was 80.1% (95% confidence interval, 76.0% to
84.2%) for unicompartmental knee replacements and 92.0% (95% confidence
interval, 90.4% to 93.6%) for total knee replacements
(Fig. 1 and Appendix). The
relative risk of revision following unicompartmental knee replacement as
compared with total knee replacement was 2.0 (95% confidence interval, 1.6 to
2.5; p < 0.001). The increased risk of revision following unicompartmental
knee replacement was seen in all age-categories (see Appendix).
Compared with total knee replacement, unicompartmental knee replacement was
associated with more revisions because of pain (relative risk, 11.3 [95%
confidence interval, 4.8 to 26.8]; p < 0.001), aseptic loosening of the
tibial component (relative risk, 1.9 [95% confidence interval, 1.2 to 3.0]; p
= 0.01) and of the femoral component (relative risk, 4.8 [95% confidence
interval, 2.3 to 10.3]; p < 0.001), and periprosthetic fracture (relative
risk, 3.2 [95% confidence interval, 1.2 to 8.9]; p = 0.02). There was,
however, a decreased risk of revision because of infection following
unicompartmental knee replacement as compared with total knee replacement
(relative risk, 0.28 [95% confidence interval, 0.10 to 0.74]; p = 0.01)
(Table II).
After ten years of follow-up, there was no significant difference in
survival among the MOD III, Genesis Uni, and Oxford II knee replacements, with
the numbers available. However, although the number of Duracon prostheses was
low (n = 47), Duracon knee replacements were associated with a significantly
higher risk of revision as compared with the other unicompartmental knee
replacements (relative risk at five years, 3.0 [95% confidence interval, 1.3
to 6.8]; p = 0.01) (Fig. 2-A
and Appendix). After five years, the Miller-Galante knee prosthesis (survival
rate, 83.0%; 95% confidence interval, 76.3% to 89.7%) had a significantly
higher rate of revision than the Oxford III knee prosthesis (survival rate,
91.1%; 95% confidence interval, 88.7% to 93.5%) (relative risk of revision,
1.8 [95% confidence interval, 1.1 to 2.8]; p = 0.01)
(Fig. 2-B and Appendix). The
results for the prostheses used primarily over the last five years were not
better than those for the prostheses used mainly the first five years of
registration, with the numbers available (see Appendix). The higher failure
rates of the Duracon and Miller-Galante knees were mainly due to more
loosening of the tibial component. For the MOD III prosthesis, there were
significantly more revisions in association with the 7.5-mm tibial components
as compared with the 9-mm components (relative risk, 3.5 [95% confidence
interval, 1.3 to 9.1]; p = 0.009). For the Genesis prosthesis, there were also
more revisions for the 8-mm tibial inserts as compared with the 10 and 12-mm
inserts (relative risk, 6.2 [95% confidence interval, 1.3 to 28.8]; p =
0.02).
Two hospitals were found to have <60% survival of the Miller-Galante
knee prosthesis after five years, while other hospitals had 100% survival
after five years (p = 0.001 [see Appendix]).
Few lateral unicompartmental knee arthroplasties were performed. For the
Genesis knee replacement, which was inserted in the lateral compartment 12% of
the time, we found no difference in survival between lateral and medial
compartment prostheses with the numbers available (relative risk, 0.53 [95%
confidence interval, 0.12 to 2.29]).
The major finding of the present study was that the rate of revision
following unicompartmental knee replacement was twice as high as that
following total knee replacement as a result of higher rates of revision due
to femoral and tibial loosening, periprosthetic fracture, and pain. The
increased risk of revision following unicompartmental knee replacement was
seen in all age-categories and had not improved during the last five-year
period as compared with the earlier period. The finding confirms those of
other register
studies14,15
but not those of studies from specialized
centers5,6,16.
The volume of surgery mattered, with the risk of revision being 40% lower
in hospitals at which twenty to forty-nine operations were performed per year
as compared with hospitals at which zero to nine operations were performed per
year. This finding corresponds well with those of registry studies on hip
replacement17,18.
For the Oxford III knee replacement, there was no difference in the risk of
revision between low and high-volume hospitals, but the duration of follow-up
was short and the number of hospitals with high numbers of procedures was low.
Our findings for the Oxford III knee replacement are in accordance with those
of more recent results from the Australian
registry19 but
contradict the findings from the Swedish
registry20 on the
earlier Oxford phase-II knee replacement, in which the results were better in
hospitals with more than twenty-three procedures per hospital per year. This
might be explained by better education of surgeons and better instrumentation
in association with Oxford III knees as compared with Oxford II knees. In the
present study, we were not able to control for the volume of operations per
surgeon, and this might have confounded our results.
Several studies have indicated that unicompartmental knee replacement is
associated with less pain and better function in comparison with total knee
replacement21,22.
In a randomized, controlled study in which the St. Georg Sled unicompartmental
knee replacement was compared with the Kinematic total knee replacement with a
patellar component, the authors found better function and less morbidity in
association with the unicompartmental knee replacement after five years and
reported the same survival for the two prostheses on the basis of the numbers
available23. These
results persisted for as long as ten
years24 and
contradict the findings of our study. Our study was an observational one, and
the patients who were selected for unicompartmental knee replacement could
have been more active and healthier and therefore could have put the
prostheses under more stress, leading to more loosening. A few dedicated
surgeons participating in a randomized study might also have performed the
more technically demanding unicompartmental knee replacements better than
surgeons across an entire country.
As we do not have information on the degree of pain and functional scores
of the patients, we cannot conclude whether the patients in the
unicompartmental knee replacement group actually had more pain than did those
in the total knee replacement group or whether the revision operation helped.
Only two of the revisions following unicompartmental knee replacement in our
study were reported as being due to progression of arthritis. Progression of
arthritis has been regarded a major cause of revision in earlier
studies7,15,16,25.
Our findings are supported by the study of St. Georg Sled unicompartmental
knee replacements, in which none of the knees progressed to
arthrosis23, and
also by the study by Berger et al., in which only one of sixty-one knees
progressed5. Our
findings could have been due to better patient selection or to a registration
bias related to the fact that progression of arthritis is not part of the
eleven standard reasons for revision on the registration form and therefore
the surgeons must mark the reason for revision as "other" and
specify progression of arthritis. Thus, some surgeons could have wrongly
marked "pain" even though the reason was progression of
arthritis.
The lower infection rate following the unicompartmental knee replacements
confirms the early results from the Swedish Knee Arthroplasty
Register26.
The Duracon prosthesis had inferior survival results than did the other
designs. The results must be interpreted with caution as the implant had been
used in a small number of patients and at only five hospitals. The Duracon
prosthesis was associated with inferior results in another
study27. The
authors concluded that the reason for the poor results was due to the use of
polyethylene that had been gamma-sterilized in air and had a shelf life of
over four years.
The Miller-Galante implant with an all-polyethylene tibial component had a
five-year survival rate of 83% in our study, which was significantly inferior
to the 91% survival rate for the Oxford III prosthesis. Berger et al. reported
a ten-year survival rate of 98% for the Miller-Galante
prosthesis5, and
Naudie et al. reported a ten-year survival rate of 90% for that
prosthesis28. In
both of those studies, a metal-backed tibial component was used, whereas in
Norway an all-polyethylene tibial component was used. In a previous two-year
study of Miller-Galante unicompartmental knee replacements, the results of
procedures performed with an all-polyethylene tibial component were similar to
those of procedures performed with a metal-backed tibial
component29,
contradicting the findings of our study. In the last yearly report from the
Swedish knee register, the Miller-Galante unicompartmental knee replacement
was reported to be associated with a higher risk for revision than the Link
unicompartmental knee replacement, but, because metal and all-polyethylene
tibial components were analyzed together, no conclusion regarding the issue of
the use of all-polyethylene tibial components can be drawn on the basis of
that report14. In a
report from Australia, the Miller-Galante unicompartmental knee replacement
was associated with good short-term results, but the report did not specify
whether an all-polyethylene or a metal-backed tibial component had been
used19. Our finding
of poor performance due to the aseptic loosening of the tibial components of
the Duracon and Miller-Galante prostheses questions the concept of using an
all-polyethylene tibial component for unicompartmental knee replacement.
As we did not register whether minimally invasive techniques were used at
the time of surgery, their influence on the individual prosthesis could not be
resolved. However, we did conduct a postal survey that was sent to all
Norwegian operating clinics in September 2003 in which we inquired about the
use of minimally invasive techniques for unicompartmental knee replacement.
The results were analyzed according to the year of surgery (beginning in 1994)
and prosthesis design. The Oxford III and the Miller-Galante prostheses were
inserted with a minimally invasive technique in 98% and 90% of cases,
respectively, and it is thus unlikely that the use of minimally invasive
techniques explains the difference in performance between these two
prostheses.
Loosening was most commonly seen in association with tibial components with
thin polyethylene, which is in accordance with the findings of other
studies30.
There was no difference in survival between lateral and medial prostheses
in cases in which the Genesis unicompartmental prosthesis had been used, which
is in accordance with the results of the Swedish knee
register25,31.
It has been recommended that a unicompartmental prosthesis should be
suitable for 20% to 30% of knee replacement
procedures6,32,
but some authors have disputed this great
percentage33. The
fact that the Oxford knee has been the most used unicompartmental knee
replacement in Norway in recent years probably explains the growing popularity
of unicompartmental knee replacement, from 5% of the total number of primary
knee replacements in 1997 to 15% in 2004. This rate is comparable with the
percentages from Sweden and
Australia14,19.
Our data were used in a decision-analysis study that showed that
unicompartmental knee replacement can be cost effective as compared with total
knee replacement in patients older than seventy years of
age34.
The procedure of unicompartmental knee replacement, especially that
involving minimally invasive techniques, has been reported to result in less
morbidity in the form of pain, faster recovery, shorter hospitalization, less
infection, less thromboembolic disease, and better range of
movement23,35,36
as compared with total knee replacement. The present study demonstrates that
these short-term advantages must be weighed against higher revision rates due
to aseptic loosening of the tibial and femoral components, persistent pain,
and periprosthetic fractures.
Additional tables and figures showing the data broken down by prosthesis
type, patient age, and hospital 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). ?