Between August 1992 and May 1996, all patients with osteoarthritis
who were scheduled to undergo a primary total knee arthroplasty
at one of two university-affiliated teaching hospitals were evaluated
for the study, which was approved by both university and hospital
ethics committees. Patients with inflammatory arthritis, a history
of patellar fracture, a prior patellectomy, patellofemoral instability,
or a prior unicondylar knee replacement were excluded. A total of
221 patients were recruited and provided informed consent. Patients
were randomized, by the use of computer-generated random numbers, to
be treated with either patellar resurfacing or patellar retention.
The randomization envelope was opened in the operating theater.
Surgery was performed by one of six experienced surgeons or their
trainees under their supervision. A Miller-Galante II prosthesis
was implanted in all patients, and all components were cemented.
When the patella was to be retained, a patelloplasty was performed.
A midline skin incision and a standard medial parapatellar approach
with preservation of the infrapatellar fat pad were used in all
patients. Optimal patellar tracking was ensured by appropriate soft-tissue
balancing, and a lateral release was performed at least 2.5 cm lateral
to the lateral patellar border if the patella subluxated during
passive range-of-motion testing. The degree of damage to the patellar
articular cartilage was recorded at the time of surgery with use
of a grading system based on that described by Outerbridge14.
Radiographs were made immediately postoperatively and annually
thereafter. At one year postoperatively, they included a full-length
weight-bearing anteroposterior view, a lateral view, and a skyline
view with 45° of knee flexion. Radiographic measurements were performed
by an independent orthopaedic surgeon.
Clinical evaluations were performed preoperatively; at three, six,
and twelve months postoperatively; and annually thereafter by one
investigator (A.J.S.) blinded to the type of treatment to which
the patient had been randomized. The Knee Society clinical rating
system15 was used at each evaluation.
Patients were carefully questioned regarding the location of any
knee pain, which was recorded as anterior, medial, lateral, posterior,
or unlocalized/general. The patient was asked to grade
the severity of the predominant area of pain using the pain scale
(of a maximum of 50 points) within the knee score of the Knee Society
clinical rating system. At the postoperative examination, the patients
were observed ascending and descending five steps of standard height.
The presence (but not the severity) of anterior knee pain, use of
the railing, and the leading limb (reciprocal, operatively treated
limb, or nonoperatively treated limb) were recorded. The test was
not timed.
Of the 221 patients originally enrolled in the study, eleven (five
with and six without patellar resurfacing) died, withdrew from the
study, or were lost to follow-up less than twelve months after the
operation and were excluded from additional analysis. In addition,
nine knees randomized to patellar resurfacing were excluded from
the study because the patella was too small for the smallest prosthesis.
Of the 201 remaining patients, nineteen had a bilateral procedure.
Each knee was randomized separately; five patients received bilateral
total knee arthroplasty without patellar resurfacing, one patient received
bilateral total knee arthroplasty with patellar resurfacing, and
thirteen patients had only one knee resurfaced. Thus, there was
a total of 220 knees, 128 without patellar resurfacing and ninety-two
with patellar resurfacing, for analysis. Nine patients died and
twelve withdrew from the study between one and three years after
the operation. These patients were included in the analysis and
treated as censored data for Cox regression analysis if the event
under investigation had not occurred. One hundred and ninety-eight
knees in 180 patients were followed clinically for three years or
more. The patients in the two treatment groups were similar with respect
to demographic variables and preoperative clinical status (Tables I and II). Thirteen patients
(four in the group without resurfacing and nine in the group with
resurfacing) had undergone a previous high tibial osteotomy.
Statistical analysis was performed with use of the SPSS for Windows
statistical package (version 8.0; SPSS, Chicago, Illinois, 1997)
or Egret (Statistics and Epidemiology Research, Seattle, Washington,
1991) in the case of Cox proportional hazards regression analysis.
Categorical data were compared with the use of the chi-square test
or the McNemar test for the comparisons of preoperative and postoperative
data. The Student t test or the Mann-Whitney U test were used to
analyze differences in continuous variables between the two treatment groups.
Nonparametric statistics were used for analysis of continuous variables
when data were not normally distributed. The Kaplan-Meier method
was used for survival estimates, and the log-rank test was used
to determine differences in survival curves. Cox proportional hazards
regression analysis was used to assess associations between potential
explanatory variables and the categorical independent variables
of patellar revision or reoperation and postoperative anterior knee
pain.
A sample size of 200 was chosen on the basis of (1) a difference
in knee scores between the two groups of 0.5 standard deviation
being of clinical relevance, and (2) an estimate of anterior knee
pain occurring in 25% of the patients without patellar
resurfacing, with a 15% effect size, significance of 5%,
and power of 90%.
Revisions and Other Reoperations
Related to the Patellofemoral Joint
Analysis of the results for all patients originally enrolled
in the study (including those lost from the study between one and
three years postoperatively) revealed that fifteen (12%)
of the 128 knees without patellar resurfacing and nine (10%)
of the ninety-two knees with patellar resurfacing underwent a revision
or another type of reoperation related to the patellofemoral articulation.
There was no significant difference, with the numbers available,
in survival rates between the replacements with and those without
patellar resurfacing when the definition of failure was a revision
or another type of reoperation related to the patellofemoral joint
(log-rank test, p = 0.739). As mentioned, the twenty-one
patients who withdrew, were lost to follow-up, or died were treated
as censored data for survival analysis if the event under investigation
had not occurred. Using Cox proportional hazards regression analysis,
we did not find that a revision or another type of reoperation related
to the patellofemoral joint could be predicted by any preoperative
or perioperative factors, including the method of patellar treatment,
lateral release, age, gender, weight, body mass index, preoperative
anterior knee pain, grade of the patellar articular cartilage and
degree of osteophytes, preoperative Knee Society knee score, and
preoperative range of motion.
Eleven of the 128 knees that had not undergone patellar resurfacing
were revised with patellar resurfacing because of anterior knee
pain, and one patient was scheduled for the procedure. Revision
resurfacing was performed at an average of twenty-five months (range,
four to forty months) after the primary operation. For the eight
patients assessed at one year or more after this procedure, satisfaction
with the result of the revision resurfacing ranged from 0% to
80%; two patients thought that the revision had resulted
in no change, five thought that it had lessened the pain, and one
patient had complete relief of pain. One additional patient underwent
arthroscopy because of anterior knee pain but had no relief and elected
not to undergo revision resurfacing. Two other patients without
patellar resurfacing underwent additional procedures because of
maltracking; one patient had a revision of the tibial component
to correct rotation with concurrent lateral release, and the other
had a tibial tubercle transfer after traumatic dislocation of the
patella.
Of the ninety-two knees that had had resurfacing of the patella,
five had a revision of the patellar component. Three of them had
the revision because of patellar loosening; one, to medialize a
laterally positioned patellar button; and one, to inset the patellar
button in an attempt to relieve anterior knee pain. A patellectomy
was performed in one patient who had sustained a comminuted fracture
of the patella in a fall. Another patient underwent lateral release
and medial reefing because of maltracking. Two patients had an arthroscopic
procedure because of anterior knee pain; one of these procedures entailed
division of a lateral capsular adhesion, and the other consisted
of lateral release, trimming of osteophytes, and manipulation under
anesthesia.
Revision Unrelated to the Patellofemoral Joint
Two (2%) of the 128 knees treated with patelloplasty
and three (3%) of the ninety-two treated with resurfacing
underwent revision or another type of reoperation for reasons unrelated
to the patellofemoral joint (loosening, infection, or removal of
loose bodies).
Anterior Knee Pain
Prevalence and Severity
All patients originally enrolled in the study (including those lost
from the study between one and three years postoperatively) were
included in the analyses of the prevalence and severity of anterior
knee pain. For the patients who underwent revision involving the
patellofemoral joint, the last follow-up examination before the
revision was considered the final data point for the analysis of
anterior knee pain. Data for analysis of anterior knee pain were
missing for one patient in each treatment group.
At the time of the latest follow-up, there was a significantly higher
incidence of anterior pain (chi square with one degree of freedom = 5.757,
p = 0.016) in the knees that had not had patellar resurfacing.
Of the 127 knees that had not had patellar resurfacing, thirty-nine
(31%; 95% confidence interval = 23% to
39%) had predominantly anterior knee pain at the time of the
latest follow-up. Of the ninety-one knees with patellar resurfacing,
fifteen (16%; 95% confidence interval = 9% to 24%)
had predominantly anterior knee pain at the time of the latest follow-up.
Of the knees that had not had patellar resurfacing, seventeen (13%)
were moderately painful anteriorly and two (2%) were severely
painful anteriorly according to the Knee Society rating system.
Only three (3%) of the ninety-one knees with patellar resurfacing
were moderately painful anteriorly, and none were severely painful
anteriorly. Patients without patellar resurfacing had significantly
worse anterior knee pain than did those with a resurfaced patella (Mann-Whitney
U test, z = -2.79, two-tailed p = 0.005). Fifteen
(12%) of the 127 knees that had had a patelloplasty and sixteen
(18%) of the ninety-one resurfaced knees were painful in
an area that was not anterior.
Temporal Development of Anterior Knee Pain
Survival analysis was performed, with any degree of anterior knee
pain present for twelve months or more as the end point (Figs. 1-A and 1-B). When anterior
knee pain had been present within the first year and had continued
for at least twelve months, its onset was recorded as occurring
on the date of the first follow-up evaluation. In some cases, this
was as early as three months postoperatively. The rate of survival
(defined by an absence of anterior knee pain) of the replacements
that included patellar resurfacing was significantly superior to
the rate for those without patellar resurfacing (log-rank test,
p = 0.020).
Predictors of Anterior Knee Pain
According to Cox regression analysis, the absence of patellar resurfacing
was the only significant predictor of anterior knee pain. Lateral
release, age, sex, weight, body mass index, preoperative anterior
knee pain, grade of patellar articular cartilage and degree of osteophytes,
preoperative Knee Society knee score, and preoperative range of
motion were not significant predictors. After adjusting for age
and sex, we found that patients without patellar resurfacing were
almost twice as likely to have anterior knee pain postoperatively
(hazard ratio = 1.95; 95% confidence interval = 1.08
to 3.54; p = 0.028).
When the two treatment groups were considered individually, the
weight of the patient was a significant predictor of anterior knee
pain in the group without patellar resurfacing (hazard ratio = 1.03;
95% confidence interval = 1.003 to 1.05; p = 0.027)
but not in the group with patellar resurfacing (p = 0.684).
After adjustment for age and sex, the influence of weight was no
longer significant (p = 0.070), although there were no
significant interaction terms between weight and age or sex. In
the group without patellar resurfacing, the mean weight of the patients
with anterior knee pain was 83.2 kg and the mean weight of those
without anterior knee pain was 77.8 kg. Interestingly, body mass
index was not related to anterior knee pain in either the resurfaced
or the nonresurfaced group (p = 0.940 and 0.776, respectively).
Knee Society Clinical Rating System Scores
and Patient Satisfaction
With the numbers available, there was no significant difference
between the knees with and those without patellar resurfacing with
regard to the Knee Society knee score (or the subscore for range
of motion) or function score or with regard to patient satisfaction.
As distributions for all three variables were negatively skewed
with several outliers in both groups, the data were analyzed with
use of the Mann-Whitney U test.
Stair-Climbing
One hundred and eighty-four patients with 202 involved knees
were interviewd at the latest follow-up examination with regard
to pain on stair-climbing. Thirty (25%) of 118 knees without
patellar resurfacing and fourteen (17%) of eighty-four
with patellar resurfacing were painful anteriorly with stair-climbing.
With the numbers available, this difference was not significant
(chi square with one degree of freedom = 2.209, p = 0.167).
One hundred and forty-three patients who had had a unilateral procedure
were assessed with regard to the leading limb while they ascended
and descended stairs. Twenty-eight (33%) of eighty-four
patients without patellar resurfacing and eleven (19%)
of fifty-eight with patellar resurfacing descended stairs one at
a time, leading with the involved limb, indicating a reluctance
or an inability to load the affected knee. One subject was unable
to descend stairs. With the numbers available, the difference between
the groups was not significant (chi square with one degree of freedom = 3.55,
p = 0.059), but it is of clinical importance.
Radiographic Findings
Skyline radiographs were available for 171 knees (77.7%); full-length,
weight-bearing, anteroposterior radiographs were available for 159
(72.3%); and lateral postoperative radiographs were available
for 174 (79.0%). Nonparametric statistical techniques were
used for analysis of radiographic measurements when scores were
not normally distributed and there were numerous outliers.
The median postoperative patellar tilt was 2° in the group with
patellar resurfacing and 4° in the group without patellar resurfacing.
This difference was significant (Mann-Whitney U test, p = 0.031).
With the numbers available, we could not detect an association between
patellar tilt and the development of postoperative anterior knee
pain (Cox regression analysis, p = 0.294). There was also
no detectable association between postoperative anterior knee pain
and the anatomical alignment or the distal femoral or proximal tibial
resection angle as measured on the anteroposterior radiographs (Cox regression
analysis, p = 0.335, 0.493, and 0.672, respectively).
The mean Insall-Salvati ratio16 was
1.1 (range, 0.63 to 1.33) in both study groups. With the numbers
available, there was no association between the development of anterior
knee pain and the Insall-Salvati ratio (p = 0.202).
A number of studies in which patellar resurfacing was performed
randomly have demonstrated that the results are not superior in
terms of pain relief4,10,13, whereas
others have shown an increased incidence of anterior knee pain after
total knee arthroplasties without patellar resurfacing5,9,12. In the present prospective,
randomized study of patients treated with the Miller-Galante II
prosthesis, the incidence of anterior pain in knees without patellar
resurfacing (31%) was significantly higher than that in
knees with patellar resurfacing (16%). In their prospective,
randomized study of the results of 118 Miller-Galante II knee arthroplasties,
Barrack et al.9 reported anterior
pain in 13% (eight) of sixty knees without patellar resurfacing
and in 7% (four) of fifty-eight knees with patellar resurfacing;
the difference was not significant. The higher incidence of anterior
knee pain in our series may be due to the fact that we used a different
system for recording knee pain and our study had a greater power
and a longer follow-up. In our series, nineteen (15%) of
127 knees without patellar resurfacing and three (3%) of
ninety-one with patellar resurfacing had predominantly anterior
pain graded as moderate or severe according to the Knee Society
clinical rating system and this difference was also significant
(chi square with one degree of freedom = 7.186, p = 0.005).
These values are comparable with those of Barrack et al. Stair descent
was better after patellar resurfacing: twenty-eight (33%)
of eighty-four patients managed with a unilateral total knee arthroplasty that
did not include patellar resurfacing, compared with eleven (19%)
of fifty-eight managed with patellar resurfacing, descended stairs
leading with the treated limb, indicating an inability or reluctance
to load the affected knee. This difference bordered on significance
(p = 0.059). The difference between treatment groups was
not as great with regard to stair ascent: fifteen (18%)
of eighty-four patients without patellar resurfacing compared with
five (9%) of fifty-eight patients with resurfacing climbed
stairs leading with the untreated limb (p = 0.111); this
finding may have been due to the fact that the knee is under less
load during stair ascent than it is during descent. Thirty (25%)
of 118 knees without patellar resurfacing were painful on stair
ascent or descent compared with fourteen (17%) of eighty-four
resurfaced knees; however, with the numbers available, this difference
was not significant. The true number of knees that were painful
while the patient ascended or descended stairs may be greater as
some patients who were unable or reluctant to load the knee stepped up
or down one stair at a time and therefore did not experience pain.
There were two reasons for the larger number of total knee arthroplasties
without patellar resurfacing than with patellar resurfacing (128
compared with ninety-two) in this series. First, blocked randomization
was not utilized in this study. Second, in nine knees randomized
to patellar resurfacing the patella was too small for the smallest
Miller-Galante II prosthesis, and these knees were excluded from
the study.
We could not identify a radiographic variable that was associated
with patellar revision or another type of patellar reoperation or
the development of anterior knee pain. However, the power for these
analyses was less than optimal as not all radiographs were available
for evaluation.
Attention to technical details and refinements of prosthetic design
appear to have substantially reduced the rate of complications of
patellar resurfacing, with recent studies demonstrating no appreciable
risk of complications compared with that associated with nonresurfacing3-8.
In our series of patients managed with the Miller-Galante II prosthesis,
the rate of revision or other reoperations related to the patellofemoral
joint was similar for the knees that had had patellar resurfacing
and those that had not (10% and 12%, respectively).
A substantial number of procedures (twelve of fifteen) in the group
originally treated with patelloplasty involved revision to resurfacing,
a relatively simple procedure, because of anterior knee pain. It
should also be noted that five knees originally treated with resurfacing
required a revision of the patellar component, a technically difficult
procedure. The average duration of follow-up in this study was forty-eight months;
problems with wear and loosening of the patellar component may increase
with time.
Resurfacing of the patella does not guarantee a painless patellofemoral
joint. Fifteen (16%) of ninety-one knees with a resurfaced
patella in our series had mild-to-moderate anterior pain, and one
revision and two investigative arthroscopic procedures were necessary
because of anterior knee pain in this group. However, the patients
with resurfacing of the patella had a significantly lower incidence
of anterior knee pain postoperatively (p = 0.016).
Patients at risk for the development of anterior knee pain need to
be identified. Of clinical note is the fact that weight was the only
preoperative variable associated with the development of postoperative
anterior knee pain in patients without patellar resurfacing, whereas
there was no relationship between body mass index and the development
of pain. This finding suggests that total joint loading, not obesity,
may be the critical factor in the development of anterior knee pain.
Several studies have shown that the design of the femoral component
influences patellofemoral contact stresses and tracking in both
resurfaced and unresurfaced patellae17-21. Results of in vitro19,20 and in vivo22,23 studies have
suggested that the optimal design for compatibility with the native
patella includes a deep trochlear groove that extends distally with
an anatomic radius of curvature. It has been shown in vitro that
the kinematics of the tibiofemoral joint can influence patellofemoral contact
force24. The results of our study
are specific to the Miller-Galante II knee prosthesis. More recently
developed prostheses designed to be more compatible with the native
patella may decrease the incidence of patellofemoral symptoms.
Note: The authors gratefully acknowledge the valuable contributions
of Mr. Greg Janes and Mr. Peter Annear, orthopaedic surgeons who
assisted in the initiation of the study, and those of Mr. Michael
Holt, Mr. Ratan Edibam, Mr. Fred Easton, Mr. John Venerys, and Mr.
Allan Wang, participating surgeons.