As described in detail in our first
article8, eighty
patients with a single symptomatic cartilage defect and no general
osteoarthritis in the knee were enrolled in this controlled randomized
multicenter study. Forty patients were treated with autologous chondrocyte
implantation and forty, with microfracture. The inclusion and exclusion
criteria and the baseline characteristics of the patients are listed in the
Appendix. The International Cartilage Repair Society (ICRS)
form9 was used to
collect demographic data and to record the history, symptoms, functional
score, pain as indicated on a visual analog scale, characteristics of the
cartilage defect, and findings of the baseline clinical examination. In
addition, the Lysholm
score10, the Tegner
score11, and the
SF-3612 were used.
Informed consent was obtained from all patients, and the study protocol was
approved by the National Review Board. Financial support was granted by the
Norwegian Ministry of Health.
The etiology of the cartilage defect was trauma in 65% of the patients and
osteochondritis dissecans in 28%; it was unknown in the rest. Most of the
defects (89%) were located on the medial femoral condyle, with the remainder
located on the lateral femoral condyle. All of the patients had a chronic
cartilage defect, with a median duration of symptoms of thirty-six months, and
93% had had knee surgery before inclusion in the study. These operations
included anterior cruciate ligament reconstruction (fifteen patients),
meniscal surgery (fourteen), arthroscopic lavage and débridement
(twenty-nine), Pridie drilling (three), and operations for osteochondritis
dissecans such as drilling or fixation of a fragment (thirteen). Autologous
chondrocyte implantation was performed according to the recommendations of the
Gothenburg
group4,13.
The conventional technique of autologous chondrocyte implantation was also
used for the osteochondral defects. No included defects were deeper than 10
mm. The technique described by Steadman et al. was used for the microfracture
procedures5,14.
Both treatment groups were managed with an identical rehabilitation protocol,
which included partial weight-bearing with crutches for eight weeks. At two
years, second-look arthroscopy with a biopsy to obtain specimens for
histological evaluation was done. An independent observer performed a
follow-up clinical examination at twelve and twenty-four months using the same
forms as had been used preoperatively.
The five-year follow-up evaluation was carried out by the first author
(G.K.), for practical and economical reasons. The operation was considered to
have failed if the patient needed a reoperation because of symptoms due to a
lack of healing of the treated defect. The need for shaving or trimming of a
lesion was not defined as a failure. As reported previously, shaving or
trimming was performed prior to the second-look arthroscopy in ten patients
(25%) treated with autologous chondrocyte implantation and four (10%) in the
microfracture
group8. The patients
with a failure remained in the study, with their last recorded clinical
follow-up scores before the failure considered to be their final clinical
score. At five years, the radiographs were classified according to the system
of Kellgren and
Lawrence15 (see
Appendix). In addition, the distance between the tibia and femur on standing
digital radiographs was measured in millimeters according to the ICRS
form9. The distances
were clustered into three groups: >4 mm, 2 to 4 mm, or <2 mm.
Statistical Methods
An original sample-size estimation showed that forty patients in each group
would be required to demonstrate a difference in the Lysholm and SF-36 scores
between the groups of at least 0.75 standard deviation from the mean, with an
alpha level of 0.05 and a power level of 90%.
T tests, the Pearson chi-square and Mann-Whitney U tests, and multiple
linear regression models were used. The level of significance was p <
0.05.
At five years, no patient had been lost to follow-up. Seven patients had
moved, were pregnant, or were not available for examination in the outpatient
clinic. However, these patients (none of whom had a failure) were contacted by
mail and/or telephone, and they returned their questionnaires. At the time of
the five-year follow-up, there were nine failures (23%) in each group compared
with two failures of the autologous chondrocyte implantation and one failure
of the microfracture treatment at two years. The failures occurred at a mean
of 26.2 months after the autologous chondrocyte implantation and 37.8 months
after the microfracture treatment (p = 0.101). The survival curve, up to sixty
months, is shown in Figure
1.
In each group, one patient with a failure had a total knee replacement and
the remaining patients with a failure were treated with a new
cartilage-resurfacing operation. Seven patients with a failure of autologous
chondrocyte implantation were treated with microfracture alone, and one
patient had, in addition to the microfracture, a high tibial osteotomy. Five
patients who had a failure of the microfracture procedure were treated with a
repeat microfracture procedure, two patients had mosaicplasty, and one patient
had an autologous chondrocyte implantation procedure.
Clinical data on the patients who did not have a failure were collected at
five years. The mean Lysholm scores and mean scores on the visual analog pain
scale remained significantly improved (p < 0.05) in both groups (Figs.
2 and
3). Compared with the baseline
values, 72% of the patients had less pain, 80% had improvement in the Lysholm
score, and 72% had improvement in the SF-36 physical component score. At two
years, we reported that more active patients, as indicated by a Tegner score
of >4 points, had significantly better clinical results (p = 0.0005). At
five years, both groups had a significantly improved mean Tegner score
compared with the baseline value: it improved from 3.28 to 4.05 points in the
group treated with autologous chondrocyte implantation (p = 0.007) and from
3.16 to 4.36 points in the microfracture group (p = 0.002). However, with the
numbers available, there was no difference between the two types of treatment
(p = 0.323).
No significant difference between the treatment groups was found in the
Lysholm score (p = 0.227) or the visual analog score (p = 0.278) at five years
after treatment, after adjustment for pretreatment measurements (with use of
linear regression analysis). At the time of the two-year follow-up, the
microfracture group had significantly more improvement in the SF-36 physical
component score compared with the patients treated with autologous chondrocyte
implantation (p = 0.004). This difference was not found at five years (p =
0.054) (Fig. 4). However,
whereas there was no significant improvement in the SF-36 physical component
score, from baseline to five years, in the group treated with autologous
chondrocyte implantation (p = 0.309), the microfracture group had a
significant improvement in the score (p < 0.001). Also, at five years
younger patients (less than thirty years old) had a better clinical outcome
than did older patients (p = 0.013), regardless of their treatment group.
At the time of the two-year follow-up, histological evaluation was
performed on biopsy specimens from sixty-seven patients (84%). There was no
significant difference between the treatment groups with regard to the
frequency with which hyaline and fibrocartilage repair tissues were found (p =
0.08). There was also no association between the clinical outcomes (according
to the Lysholm score [p = 0.160], the visual analog scale [p = 0.175], and the
SF-36 physical component score [p = 0.850]) at five years and the histological
quality of the specimens (as assessed with a semiquantitative grading system
ranging from 1 [best] to 4 [worst]) at two years. A comparison of the
histological quality of biopsy specimens from twelve patients with a failure
with that of fifty-five patients without a failure revealed that none of the
failures were in knees with the best cartilage-quality score (p = 0.001).
However, comparison of the histological quality scores of patients with and
without failure with use of cross-tabulation showed no general significant
difference (p = 0.118) (Table
I). We found an association between the quality of the repair seen
macroscopically (recorded at the arthroscopy done at two years) and the risk
of treatment failure (p = 0.020). A good repair according to the macroscopic
appearance reduced the risk of treatment failure, and a bad repair increased
the risk of failure.
Twenty-four percent (fifteen) of the sixty-two patients in the study in
whom the procedure did not fail demonstrated a reduced joint space (<4 mm)
on standing knee radiographs and 34% (twenty-one) of the patients had at least
grade-2 changes according to the system of Kellgren and Lawrence at five
years. However, with the numbers studied, no significant difference was found
between the group treated with autologous chondrocyte implantation and the
microfracture group regarding the frequency of radiographic changes at five
years.
We did find an association between osteoarthritis, as measured on the
Kellgren and Lawrence scale, and pain, as measured on the visual analog scale,
at five years (p = 0.035). For this regression analysis calculation, the
five-year visual analog score was used as the dependent variable and the
radiographic grade was used as a constant predictor. This association suggests
that patients with pain in the knee were more likely to have radiographic
signs of early osteoarthritis. The same association was found between the
SF-36 physical component score and radiographic evidence of osteoarthritis
when the distance between the femur and tibia was measured (p = 0.026). With
the numbers available, we did not find a significant association between the
Lysholm score and radiographic evidence of osteoarthritis, as demonstrated by
the distance between the femur and tibia (p = 0.208) or according to the scale
of Kellgren and Lawrence (p = 0.294).
In this prospective randomized study of relatively large chronic cartilage
defects, both autologous chondrocyte implantation and microfracture resulted
in significant clinical improvement at five years after treatment. Seventy-two
percent of the patients had less pain and 80% of the patients had a better
Lysholm score five years after treatment compared with preoperatively.
However, there was a 23% rate of treatment failure in each group.
Autologous chondrocyte implantation and microfracture yielded similar
results at two and five years. There was no significant difference between the
groups regarding the clinical scores or failure rate at five years. A major
strength of the study is that no patients were lost to follow-up, although
some were only contacted by telephone and/or mail. This is also one of the
largest, most comprehensive studies of these two procedures, which included
evaluation of clinical results, histological examination of biopsy specimens
from cartilage repair sites at two years, macroscopic results, and radiographs
of patients randomized to treatment with autologous chondrocyte implantation
or microfracture. A relative weakness of the study is that, unlike the
two-year follow-up evaluation, which was performed by an independent observer,
the five-year follow-up evaluation was carried out by an author of the study.
However, the risk of bias was reduced by using patient-based scoring
forms.
Autologous chondrocyte implantation is a more technically demanding
procedure than microfracture; it also requires two separate operations,
including an arthrotomy for the implantation of the cells. However, all
surgeons were experienced knee surgeons trained in both procedures. Our
rehabilitation protocol was in line with
recommendations4,5.
As was the case in other published
reports4,5,7,
we have no data on patient compliance with the rehabilitation protocol. The
location of the defect may also be an important factor. Recently,
microfracture was found to have less favorable results when it was used to
treat patellofemoral lesions, and autologous chondrocyte implantation may be a
better option for trochlear
defects16. Most
(89%) of the defects in this study were located on the medial femoral condyle,
with the remaining ones on the lateral femoral condyle. Finally, our cohort of
patients had chronic and relatively large lesions. No patient with an acute
injury was included.
Another limitation of our study is the lack of a control group that was not
treated with surgery or was treated simply with arthroscopic lavage. Our
patients had experienced pain for a long period of time, and the majority had
undergone previous surgery and conservative treatment. Consequently, it would
have been difficult to assign patients to such control groups.
The natural history of cartilage injuries is still not well understood, and
we do not know which patients will benefit from surgical interventions. Linden
performed a long-term follow-up study of patients with osteochondritis
dissecans of the femoral
condyles17. He
concluded that patients with adultonset osteochondritis dissecans had
osteoarthritis ten years earlier than might be expected for the onset of
primary osteoarthritis. In contrast, when the osteochondritis dissecans was
diagnosed in childhood, the patient had no increased risk of osteoarthritis
later in life when compared with the normal population. Shelbourne et al.
found that chondral injuries noted at the time of anterior cruciate ligament
reconstruction did not influence the clinical outcome of the operation at a
mean of 8.7
years18. However,
the defects were small, and the population was young. In contrast to the
observations in that study, two of us (J.O.D. and T.G.) found a significant
relationship between cartilage injury detected prior to anterior cruciate
ligament reconstruction and osteoarthritis found eight years postoperatively
(p <
0.005)19.
In our study, we included only chronic symptomatic lesions (median duration
of symptoms, thirty-six months). Acute chondral and subchondral injuries may
heal completely or partially with repair tissue and become asymptomatic
regardless of treatment, and this makes it difficult to include these injuries
in a controlled study. Some studies have included patients with acute lesions.
For example, 21% of the lesions were acute in the evaluation of microfracture
by Steadman et al.5.
It is difficult to understand the mechanism of chronic pain and disability.
Even though cartilage defects are found, they can be asymptomatic and the pain
may be caused by other conditions in the knee or the surrounding structures.
Referred pain and a psychosomatic influence also have to be considered, but
these aspects are not easily evaluated in a study such as the present one.
Bodily pain lasting for more than six months is often classified as
chronic20.
In our trial, there was a 23% failure rate in each treatment group by the
five-year follow-up time-point. Peterson et al. reported a failure rate of 11%
after autologous chondrocyte implantation on the femoral condyles, with most
of the failures occurring less than two years
postoperatively13.
Their clinical success rate has been quoted to be from 80% to 90%, and they
concluded that a graft surviving for two years is likely to remain viable
three to eight years later. In a recent study of the clinical outcomes of
autologous chondrocyte implantation in the United States (data derived from
the Carticel prospective cartilage repair registry), a five-year follow-up was
completed for eighty-seven of 100
patients21. The
treatment failed in thirteen patients (15%), and overall sixty-two patients
had improved scores at the time of follow-up. However, the results for the
thirteen missing patients are unknown. The mean defect size in that study was
4.9 cm2, which is comparable with the defect size in our study.
Success rates similar to those reported for autologous chondrocyte
implantation have been reported after microfracture procedures. Steadman et
al. found that, at seven years after microfracture surgery, 80% of their
patients rated their condition as
improved5. All of
these patients were less than forty-five years of age when they were enrolled
in the study, and the mean size of the defects (2.77 cm2) was
smaller than it was in our study. That study also included some acute defects.
Recently, Kreuz et al. reported good short-term results following
microfracture, although they observed deterioration of some of those results
starting at eighteen months after the
surgery16.
Horas et al. presented the results of a prospective, comparative trial of
the two-year outcomes of forty patients randomized to be treated with either
autologous chondrocyte implantation or osteochondral cylinder
transplantation22.
Each group consisted of twenty patients. No significant difference in the
Tegner and Meyers scores was found between groups, although the Lysholm scores
at six, twelve, and twenty-four months were significantly higher in the group
treated with osteochondral cylinder transplantation (p <0.05).
Bentley et al. reported that autologous chondrocyte implantation yielded
better results than did mosaicplasty, although the difference was not
significant in the overall series, and they found hyaline cartilage in seven
of nineteen biopsy specimens obtained at one year after autologous chondrocyte
implantation23. In
contrast to many rehabilitation protocols, the protocol in that study included
immobilization in a cylinder cast for ten days after the surgery. Subgroup
analysis showed that the Cincinnati score was significantly better for
patients in whom the medial femoral condyle had been treated by autologous
chondrocyte implantation than it was for those who had mosaicplasty at that
location (p < 0.032).
The above
studies22,23,
in addition to the one from our
group8, were three
of the four studies included in the recent Cochrane
review6. The fourth
study included in that review was a non-randomized study comparing
twenty-seven matrix-guided autologous chondrocyte implantation procedures with
seven microfracture
operations24. The
authors reported significantly better Lysholm scores at twenty-four months in
the group treated with matrix-guided autologous chondrocyte implantation (p =
0.049), but differences in the Tegner and ICRS scores at twenty-four months
were not significant.
The preliminary results of a clinical study comparing periosteal covering
of femoral defects in the knee with and without implantation of cultured
autologous chondrocytes was presented at the ICRS conference in San Diego in
200625. No
clinically relevant difference between the two groups was found at twelve
months. However, biopsies done at the same time-point showed significantly
more hyaline cartilage in the group that received chondrocytes (p =
0.003).
In our study, 34% of the patients had radiographic evidence of early
osteoarthritis at five years after treatment. This is remarkable considering
that, at this time-point, our patients were in their late thirties and the
fact that radiographic evidence of osteoarthritis was an exclusion criterion
at the outset of the study. An unfortunate weakness of our original study
design, however, was that the Kellgren and Lawrence methodology was not used.
The standing knee radiographs were merely graded as normal by the radiologist
if there was at least a 4-mm distance between the tibia and femur. Lohmander
et al. found that 51% of female soccer players (mean age at assessment,
thirty-one years) had radiographic changes (comparable with a Kellgren and
Lawrence grade of 2) in the knee twelve years after injury to the anterior
cruciate
ligament26. At
baseline, 93% of our patients had had previous surgery. Thus, considering the
relatively large and chronic defects that were included in our study, it may
not be surprising that early radiographic signs of arthritis were present in
about one-third of the patients at the time of the five-year follow-up. As
described by Englund, several studies have shown only a limited correlation
between radiographic signs of osteoarthritis and clinical outcome scores, and
he reported that nearly half of patients who had radiographic evidence of
osteoarthritis of the knee were
asymptomatic27.
However, we found an association between pain and radiographic evidence of
osteoarthritis as defined with the Kellgren and Lawrence scale, and this has
also been reported by
others28,29.
A comparison of the histological quality of the repair tissue between
patients with and without failure of treatment revealed no significant
difference, but none of the patients with a failure in our study had the
best-quality cartilage. This finding suggests that repair cartilage, which is
predominantly hyaline, at two years may reduce the risk of later failure.
The results of our study led us to propose that microfracture, a low-cost
and minimally invasive procedure, should be preferred as the first-line
cartilage repair procedure for defects located on the medial or lateral
femoral condyle of the knee. Autologous chondrocyte implantation may be
preferred as a second-line treatment, particularly for large defects that are
not contained.
Further research is needed to improve current surgical treatment of
cartilage defects. There must be improvements in surgical techniques as well
as in the field of cellular and molecular biology. Valid clinical answers in
this field can be obtained only through the combination of additional
randomized clinical trials and strictly controlled independent registry
reviews.
Tables showing the inclusion and exclusion criteria and the basic patient
characteristics of the original study as well as the Kellgren and Lawrence
grading system 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). ?