After giving informed consent for study participation, and following the
initial assessment, the subjects were randomized to one of two study groups:
primary total knee arthroplasty with a hydroxyapatite-coated tibial component
or primary total knee arthroplasty with cemented tibial fixation. The
randomization sequence was computer-generated in blocks of twenty subjects,
and the randomization codes were stored in sequentially numbered opaque
envelopes. The study received ethical approval from the regional institutional
review board.
The study was powered to detect a 10-point difference between groups on
either the functional or the pain scale of the Western Ontario and McMaster
Universities Osteoarthritis Index (WOMAC) (a = 0.05, ß = 0.20,
power = 80%)28.
The subjects were recruited during their preoperative assessment at
participating institutions from November 1996 to August 2000. Ten orthopaedic
surgeons practicing at three sites contributed cases to the study; all sites
are tertiary healthcare centers in one Canadian health region.
Patients were eligible for the study if they were scheduled to undergo a
primary total knee arthroplasty to treat noninflammatory arthritis, were less
than seventy years old, were able to speak and read the English language or
had an available translator, and were willing to return for follow-up visits.
Subjects were excluded if they had a history of knee infection, a previous
patellectomy or high tibial osteotomy, a deficient posterior cruciate
ligament, a knee flexion contracture or a varusvalgus deformity of
>20°, <90° of knee flexion, or tibial or femoral bone deficiency
that would require augmentation.
A standard surgical technique involving a midline incision and medial
parapatellar exposure was utilized by all surgeons. Because the study centers
are teaching hospitals, orthopaedic residents were involved in some of the
operations, under the direct supervision of the participating surgeons. A
standardized regional clinical pathway was in place throughout the study,
ensuring that all patients received similar preoperative, perioperative, and
postoperative care. All patients bore weight as tolerated using walking aids
for the first six postoperative weeks.
All study prostheses were manufactured by Stryker Orthopaedics (Mahwah, New
Jersey). All subjects received an uncemented femoral component without a
hydroxyapatite coating. Patellar resurfacing was done at the surgeons'
discretion, and all patellar components were cemented.
Subjects randomized to the cement group received a Scorpio Series-7000
standard tibial tray made of cobalt-chromium alloy with a light bead-blasted
and satin finish in a waffle pattern at the bone interface. The cemented
implant was fixed with polymethylmethacrylate cement without embedded
antibiotics.
Subjects randomized to the hydroxyapatite group received a Scorpio
hydroxyapatite-coated fixed baseplate made of cobalt-chromium alloy with a
light bead-blasted and satin finish (3.0 to 4.3 µm Ra) in a waffle pattern
at the bone interface. The hydroxyapatite coating was heated and applied by
plasma spray with a minimum of 90% (by weight) of the particles <100 µm
in diameter, an in-house-generated powder density of =3.00
g/cm3, and a thickness of 40 to 70 µm.
The operative and perioperative measures that were recorded included the
type of anesthesia, the operating time (from incision to wound closure), and
any adverse events related to the surgery or occurring in the immediate
postoperative period.
During evaluation in the preadmission clinic, a physical therapist assessed
the range of motion of the knee and the subject completed the WOMAC and RAND
36-Item Short Form Health Survey (RAND-36). The subjects were randomized after
the initial assessment, so the evaluator was blinded to the group
allocation.
Of the eighty-one subjects enrolled, forty-one were randomly allocated to
the cement group and forty, to the hydroxyapatite group. The participating
surgeons, including those who supervised orthopaedic residents, contributed
similar numbers of subjects to each group (p = 0.31), which should have
prevented surgeons from being a confounding variable. No differences were seen
in baseline or demographic characteristics (for example, age or gender)
between groups (Table I). There
were also no differences in operating-room time or type of anesthesia
(Table I). Most subjects did
not have the patella resurfaced, and the numbers of subjects who did were
similar in the two groups (Table
I).
Subjects were reassessed, independent of the orthopaedic surgeon who
performed the operation, at six months, one year, and five years after surgery
by physical therapists not involved in the subjects' clinical care. The
subjects completed the same questionnaires as they had completed
preoperatively, and the physical therapist performed a knee assessment. The
subjects were asked about complications and reoperations at each follow-up
visit. Anteroposterior and lateral radiographs were made at each visit unless
they had been obtained in the previous six months.
Outcome Measures
The WOMAC, shown to be valid, reliable, and sensitive to changes in pain
and function in patients with osteoarthritis treated with a total knee
arthroplasty, was used to determine patient-reported joint-specific pain and
function1,29-31.
Each WOMAC subscale score was transformed to a range of 0 to 100 points, with
a score of 100 points indicating no pain or dysfunction, so that the RAND-36
and WOMAC scores were
unidirectional32.
The RAND-36 was used to determine overall health status. The RAND-36 is a
thirty-six-item general health questionnaire that contains items identical to
those in the Short Form-36 (SF-36). Eight dimensions of health are measured:
physical functioning, physical role functioning, bodily pain, general health
perception, vitality, emotional role functioning, social functioning, and
mental
health33-35.
The validity and reliability of the questionnaire have been tested in
populations of patients treated with joint
arthroplasty1,30,32,36,37.
The active range of motion of the knee was assessed, with the patient
supine, by a physical therapist utilizing a goniometer, a method that has been
reported to be reliable and valid for knee range-of-motion
measurement38.
Anteroposterior and lateral radiographs were made following the surgery and
at each postoperative visit. Radiolucencies were rated with use of the Knee
Society radiographic rating
scale39 by an
experienced arthroplasty surgeon who had not performed any of the operations
in the study. Progression of radiolucent lines was defined as at least a 1-mm
increase in the length or width of the line since the previous evaluation.
Analysis
All of the analyses were performed according to the intention-to-treat
principle, in which patients' data are analyzed on the basis of their group
allocation rather than the treatment that was actually received. Descriptive
statistics (means, standard deviations, and proportions) were generated for
all study variables. Potential systematic differences at baseline, including
surgical characteristics that could impact outcomes, were evaluated with use
of standard bivariate tests (independent t tests and chi-square tests).
Self-reported pain and function scores were the primary outcome variables.
The WOMAC scale, which measures joint-specific pain and function, and the
RAND-36 dimensions of bodily pain and physical function were used to compare
pain and function over time between the two groups. The primary analysis of
these variables was a two-way repeated-measures analysis of variance.
Secondarily, we performed independent t tests at each assessment period when
there was a =9-point difference between groups to determine if these
differences were significant, as 9 points had been defined a priori as the
minimum for a clinically relevant
difference28.
Imputation of the data was performed for subjects who missed either the
six-month or the one-year assessment, to maintain the subject's data in the
repeated-measures analysis. The imputed value consisted of the group mean for
each variable specific to the follow-up period. For example, a subject in the
cement group who missed only the six-month follow-up assessment was assigned
the cement group's six-month mean for each study outcome measure as an imputed
value for the six-month assessment. Sensitivity analysis was undertaken with
imputed values included and then excluded to determine the effect of
imputation on the results. Subjects who missed the five-year follow-up
assessment were considered lost to follow-up. Subjects who missed both the
six-month and the one-year follow-up visits were not included in the
repeated-measures analysis.
We compared the other six dimensions of the RAND-36 over time to determine
if any other health dimensions were affected by the intervention. The ranges
of motion of the knee were also compared to determine if the method of tibial
fixation affected knee movement.
Complication rates, revision rates, and radiographic data were compared
between the two groups with use of nonparametric chi-square tests. For
radiographic data, counts of the total numbers of lucencies in the defined
zones39 within six
months after the surgery, and again at one and five years following the
surgery, were compared between the groups. The subjects were also rated as
having progressive changes or having a stable radiographic appearance by
comparing radiographs over time.
All statistical analyses were performed with the Statistical Package for
the Social Sciences (SPSS), version 14.0 (SPSS, Chicago, Illinois), with use
of two-tailed tests and a significance level of a = 0.05.
Of the eighty-one subjects, seventy (86%) completed the five-year
assessment (Fig. 1). One
subject died in the hospital as a result of postoperative medical
complications that were not related to the intervention. Ten subjects withdrew
from the trial: four did so prior to the six-month evaluation; three, prior to
the one-year evaluation; and three, prior to the five-year evaluation.
Six-month scores had to be imputed for eight subjects, whereas one-year scores
had to be imputed for only three subjects. Three subjects who missed the
six-month and one-year assessments, but returned for the five-year assessment,
were not included in the repeated-measures analyses. Subjects requiring
imputation of values were not significantly different in terms of
self-reported pain or function preoperatively from those who had returned for
all follow-up assessments (p > 0.05). As imputation did not impact the
results, all results of the repeated-measures analyses reflect those that
included the imputed values, to maximize use of available data.
The self-reported pain assessments differed slightly between the groups at
the six-month evaluation, but the two groups reported similar pain levels at
all of the other evaluations (Table
II). Although there were ten participating surgeons, we found no
differences among these different surgeons' patients with regard to changes in
self-reported pain (p > 0.05).
All subjects reported much decreased pain levels as measured with the WOMAC
pain index at all postoperative assessments. The only postoperative time
period at which group differences approached 10 points—the minimum for a
clinically important difference— was the six-month period, when the
hydroxyapatite group reported more pain
(Table II). The group
difference in WOMAC pain scores at six months was not significant when it was
analyzed with an independent t test (p = 0.11). A similar pattern was noted
for the bodily pain dimension of the RAND-36, with the hydroxyapatite group
reporting more pain at the six-month postoperative assessment only
(Table II). This difference was
found to be significant when it was analyzed with an independent t test (p =
0.006).
The groups were similar with regard to self-reported knee function prior to
the surgery, with both indicating that they experienced marked limitation in
daily activities that was related to the knee condition. Postoperatively, both
groups reported significant functional improvement, with no differences noted
between the groups (Table
III).
At no assessment period was there a substantial difference between groups
in terms of joint-specific function as measured with the WOMAC functional
scale (Table III). Patients in
both groups reported significantly improved function by six months
postoperatively, with very little change from that level noted at either the
one-year or the five-year assessment.
The physical function dimension of the RAND-36 also demonstrated no
significant difference between groups
(Table III). On secondary
analyses, the lower score for physical function reported by the hydroxyapatite
group was not significantly different from the score in the cement group at
either the six-month (p = 0.39) or the one-year assessment (p = 0.36).
None of the remaining six dimensions of the RAND-36 differed between the
groups at any assessment (p > 0.05 for group effect). The subjects in the
two groups reported similar limitations preoperatively and similar
improvements postoperatively in the physical health dimensions of physical
role functioning (p < 0.001 for time effect) and general health (p = 0.035
for time effect). Of the dimensions considered more relevant to the patients'
mental health, only social functioning improved significantly over time (p
< 0.001); the scores for emotional role functioning, vitality, and mental
health did not change significantly in either group.
Prior to the surgery, the active range of knee motion (flexion plus
extension) averaged 101.9° ± 17.9° in the cement group compared
with 104.1° ± 19.8° in the hydroxyapatite group (p = 0.60).
Postoperatively, there were no group differences in the active range of motion
over time, with 100.3° ± 19.9° reported for the cement group
and 103.1° ± 13.2° reported for the hydroxyapatite group at the
five-year assessment (p = 0.64).
No significant differences in the findings on plain radiographs were seen
between the groups over the five-year study interval (p = 0.29). Of the
seventy subjects who returned for the five-year assessment, fifty-five (79%)
had five-year radiographs available for review. Eighty-four percent
(forty-six) of these fifty-five subjects had no lucent line or only one lucent
line of =1 mm detected at the five-year assessment. Three subjects, one
with hydroxyapatite fixation and two with cement fixation, appeared to have
progressive loosening over the five years, as evidenced by increasing
lucencies in multiple zones. All of those subjects reported increasing knee
pain between the one year and five-year assessment periods, two of them (one
with hydroxyapatite fixation and one with cement fixation) also reported
deteriorating knee function between one and five years.
Of the patients who survived to hospital discharge, 86% (sixty-nine)
reported no postoperative or postdischarge complications, with no differences
noted between groups (p = 0.65). Three subjects in each group had
postoperative wound redness or cellulitis, which resolved with oral
administration of antibiotics. Three subjects (one in the cement group and two
in the hydroxyapatite group) required manipulations within the first six
postoperative months because of poor knee flexion. One subject in the
hydroxyapatite group required revision to a thicker liner three years after
the surgery, and one subject in the cement group required a patellar revision
two years after the initial surgery.
There were no major differences between the two groups over the initial
five-year postoperative period. Both fixation methods resulted in good
patient-reported, clinical, and gross radiographic outcomes during our
evaluation period, and no patient underwent revision of the tibial component
before the five-year assessment.
However, it is interesting to note that, at the six-month assessment
period, there was a mild difference in self-reported pain as shown by two
validated outcome measures that are responsive to change following total knee
arthroplasty1,36.
We are not aware of any other studies in which patient-reported
condition-specific measures were used or that focused on clinical outcomes
during the early postoperative
period12-14,16,19.
The difference that we found during the postoperative recovery period could be
related to the initial prosthetic migration that has been reported by others
who have performed more in-depth radiographic
studies12,13,16,19.
The first six months may represent the time needed for biological fixation to
occur12,16;
thus, it is possible that there were greater differences in pain prior to the
six-month evaluation. Our finding should be studied in more detail in future
investigations of biological fixation in the tibia, with particular emphasis
on the early postoperative period. However, on the basis of our findings,
clinicians should not be alarmed if patients initially report slightly higher
pain levels with cementless fixation, as the reported difference in this study
disappeared by the one-year assessment and did not affect the five-year
results.
We detected no advantages to biological fixation; however, five-year
outcomes can be considered only medium-term results, so future evaluation of
these patients is required to determine if there are any long-term advantages
or disadvantages to cementless fixation with hydroxyapatite. The
hydroxyapatite-coated prostheses were three times more expensive than the
cemented tibial prostheses. However, the cost of the cement would increase the
cost of the cemented tibial fixation. Operative time was also similar between
the groups, suggesting that there is no cost savings in terms of
operating-room utilization. As we did not perform a formal cost analysis, we
can only state the cost differentials between the prostheses themselves.
Future investigators should consider the impact of cost in relation to
outcomes as part of their analysis plan.
The operations in our study were performed by both high and low-volume
arthroplasty surgeons. No differences were detected between their results,
suggesting that similar outcomes can be expected by surgeons who perform a
large number of joint arthroplasties (more than fifty) each year and those who
perform fewer operations (less than thirty) per year. Surgeons were also
similarly distributed across both groups, as was expected in a randomized
study, reducing the likelihood of surgeon experience being a confounding
variable. Still, we should note that all operations were performed at three
high-volume tertiary health-care centers that cumulatively perform more than
1500 total knee arthroplasties each year. Others have suggested that
performance of arthroplasties in a high-volume center reduces the risk of
postoperative complications regardless of surgeon
volume40. We also
excluded patients requiring complex surgical procedures because of
preoperative deformity or inflammatory joint disease, so our findings may not
apply to those patient groups.
The femoral components used in both groups were cementless; thus, the early
difference seen between groups cannot be attributed to pain arising from the
quality of the femoral fixation. If cementless femoral fixation was the only
source of pain, we would not have detected any group differences. Furthermore,
patellar resurfacing was similarly distributed across the two groups, reducing
the likelihood that the patellar treatment was a source for differences in
pain between groups.
The patients were assessed independently by physical therapists who had not
provided the subjects' clinical care and who were blinded to group allocation.
We applied validated, reliable, and responsive outcome measures at regular
intervals throughout the first five postoperative years, and the study
included a prospective preoperative evaluation. The follow-up rate was also
acceptable as seventy (86%) of the patients were available for the final
five-year assessment.
Our radiographic findings are limited as we had only plain radiographs with
which to evaluate implant migration and fixation, and five-year radiographs
were not available for all of the subjects. Authors of other studies who have
utilized fluoroscopy and radiostereometric analyses have been able to detect
minimal radiographic differences between
groups12,13,16,19.
We could detect only gross radiographic changes over the five-year assessment
period; very few occurred, and there were no group differences.
In summary, we cannot recommend or discourage the use of cementless tibial
fixation with hydroxyapatite instead of cemented tibial fixation, as patients
reported similar five-year clinical outcomes and similar results were seen on
plain radiographs. Although patients with hydroxyapatite fixation reported
slightly increased pain in the first six months following the surgery, this
difference disappeared by one year and did not negatively impact the five-year
results. As our results can only be considered medium-term, longer-term
evaluation is required to determine if cementless tibial fixation with
hydroxyapatite offers any advantages with regard to the longevity of the
implant or reduces the complexity of revision, when required, in comparison
with cemented tibial fixation. ?