Intraarticular location predicts cartilage filling and subchondral bone changes in a chondral defect

Background and purpose The natural history of, and predictive factors for outcome of cartilage restoration in chondral defects are poorly understood. We investigated the natural history of cartilage filling subchondral bone changes, comparing defects at two locations in the rabbit knee. Animals and methods In New Zealand rabbits aged 22 weeks, a 4-mm pure chondral defect (ICRS grade 3b) was created in the patella of one knee and in the medial femoral condyle of the other. A stereo microscope was used to optimize the preparation of the defects. The animals were killed 12, 24, and 36 weeks after surgery. Defect filling and the density of subchondral mineralized tissue was estimated using Analysis Pro software on micrographed histological sections. Results The mean filling of the patellar defects was more than twice that of the medial femoral condylar defects at 24 and 36 weeks of follow-up. There was a statistically significant increase in filling from 24 to 36 weeks after surgery at both locations. The density of subchondral mineralized tissue beneath the defects subsided with time in the patellas, in contrast to the density in the medial femoral condyles, which remained unchanged. Interpretation The intraarticular location is a predictive factor for spontaneous filling and subchondral bone changes of chondral defects corresponding to ICRS grade 3b. Disregarding location, the spontaneous filling increased with long-term follow-up. This should be considered when evaluating aspects of cartilage restoration.

Background and purpose The natural history of, and predictive factors for outcome of cartilage restoration in chondral defects are poorly understood. We investigated the natural history of cartilage filling subchondral bone changes, comparing defects at two locations in the rabbit knee.
Animals and methods In New Zealand rabbits aged 22 weeks, a 4-mm pure chondral defect (ICRS grade 3b) was created in the patella of one knee and in the medial femoral condyle of the other. A stereo microscope was used to optimize the preparation of the defects. The animals were killed 12, 24, and 36 weeks after surgery. Defect filling and the density of subchondral mineralized tissue was estimated using Analysis Pro software on micrographed histological sections.
Results The mean filling of the patellar defects was more than twice that of the medial femoral condylar defects at 24 and 36 weeks of follow-up. There was a statistically significant increase in filling from 24 to 36 weeks after surgery at both locations.
The density of subchondral mineralized tissue beneath the defects subsided with time in the patellas, in contrast to the density in the medial femoral condyles, which remained unchanged.
Interpretation The intraarticular location is a predictive factor for spontaneous filling and subchondral bone changes of chondral defects corresponding to ICRS grade 3b. Disregarding location, the spontaneous filling increased with long-term follow-up. This should be considered when evaluating aspects of cartilage restoration.  Focal articular cartilage injuries of the knee are common (Hjelle et al. 2002, Aroen et al. 2004) and they can impair patients' quality of life as much as severe osteoarthritis . The literature concerning the natural history of focal cartilage defects in patients, and the intrinsic factors affect� ing it, is limited (Linden 1977, Messner and Gillquist 1996, Drogset and Grontvedt 2002, Shelbourne et al. 2003. �n e�perimental studies evaluating cartilage restora� . �n e�perimental studies evaluating cartilage restora� �n e�perimental studies evaluating cartilage restora� tion in general, the importance of intrinsic factors such as the depth and size of the lesion and the time from when the lesion was made to evaluation have been emphasized (Shapiro et al. 1993, Hunziker 1999, Lietman et al. 2002. Which part of the joint is affected and whether or not the defect is weight�bear� ing are also of interest (Hurtig 1988, Frisbie et al. 1999. Most of these studies have, however, concerned defects penetrating the subchondral mineralized tissues corresponding to �CRS grade 4 (Brittberg and Winalski 2003). Access to bone marrow elements in these defects might be one of the strongest predic� tive factors for filling of the defect, making the importance of other factors difficult to evaluate (Hunziker 1999).
�n e�perimental studies on pure chondral defects that do not penetrate the subchondral mineralized tissues, corre� sponding to �CRS grade 3b (Brittberg and Winalski 2003), the type of animal studied, the size of the lesion, and the location of the defects vary, and there is limited data on the influence of these parameters on outcome (Breinan et al. 2000). The information on spontaneous filling comes mainly from observations of untreated defects serving as controls (Grande et al. 1989, Brittberg et al. 1996, Breinan et al. 1997, 2000, Frisbie et al. 1999, Dorotka et al. 2005) and the information on subchondral bone changes is even more limited (Breinan et al. 1997, Frisbie et al. 1999. Although most human focal cartilage lesions are located on the medial femur condyle (Aroen et al. 2004), there have been few e�perimental studies involving untreated �CRS grade 3b defects on the medial femur condyle (Dorotka et al. 2005). According to a PubMed search, the rabbit knee is the most widely used e�perimental animal model for cartilage restora� tion (Årøen 2005). The locations of �CRS grade 3 chondral defects in the rabbit knee evaluated for spontaneous changes have included the patella (Grande et al. 1989, Brittberg et al. 1996) and, in one study, defects at the distal surface of the femur (Mitchell and Shepard 1976). The latter report did not, however, include quantitative data.
To our knowledge, the influence of the intraarticular loca� tion on the outcome of cartilage restoration and subchondral bone changes has not been thoroughly studied. Thus, the main purpose of our study was to test the hypothesis that the intraarticular location influences the spontaneous filling of a chondral defect that does not penetrate the subchondral bone. Secondly, we wanted to evaluate whether the intraarticular location would influence changes in the subchondral bone and degenerative changes as evaluated from macroscopic appear� ance and proteoglycan content of synovial fluid (Messner et al. 1993a).

Methods
We used our established e�perimental animal model (Aroen et al. 2005). Adult �ew �ealand rabbits were included in a ran� . Adult �ew �ealand rabbits were included in a ran� domized study where circular lesions 4 mm in diameter were created in the patella of one knee and compared in pairwise fashion to identical lesions in the medial femoral condyle of the contralateral knee. The lesions were pure chondral-down to, but not penetrating the calcified layer-corresponding to �CRS grade 3b. Follow�up was 12, 24, and 36 weeks after ini� tial surgery.
The main endpoint was difference in degree of tissue filling between the defects in the patella and the condyles at each fol� low�up. The secondary endpoints were difference in changes in density of subchondral mineralized tissue with time and difference in degenerative changes, evaluated by macroscopic appearance and joint fluid proteoglycan content, between the two locations at each follow�up.

Care of animals
The conditions for the animals, their diet, the use of ster� ile conditions perioperatively, capture of synovial fluid for proteoglycan analyses, anesthesia, analgesia, and the killing procedure have all been described previously (Aroen et al. 2005). Throughout the follow�up period, the e�perimental animals gained weight and achieved a weight at follow�up that was similar to that of control animals of corresponding age ( Figure 1). The animals were allowed to move freely in their cages, and all of them were able to bear weight on both e�tremities immediately after surgery. The surgeons were e�perienced orthopedic surgeons and they were certi� fied according to the rules of animal care and e�perimental surgery. The study was performed according to the guide� lines for animal research at the University of Oslo and it was approved by the Committee for E�perimental Animal Care of the �orwegian government.
Experimental groups 41 adult �ew �ealand rabbits were used. The animals had defects created in both knees at the age of 22 weeks, in the patella of one and in the medial femoral condyle of the other. To avoid the possibility of the "learning curve" being a form of bias, the animals were block�randomized for killing at the follow�up time points. 12 animals were planned to be killed at 12 weeks, 12 at 24 weeks, and 17 at 36 weeks. However, due to substantial loss of animals during follow�up, the decision was made to spare animals to ensure that there was a sufficient number at 36 weeks of follow�up, since this was considered to be the most important follow�up time point of the study. Partly because of this shift in group sizes and partly because of other complications during the study, the final numbers of patellas and condyles available for evaluation were 8 and 8, respectively, at 12 weeks, 9 and 9 at 24 weeks, and 17 and 18 at 36 weeks of follow�up. The numbers of animals available for paired analysis (with sections from both knees available for evaluation) were 8 at 12 weeks, 7 at 24 weeks, and 17 at 36 weeks. �n addition, 20 knees from 10 animals with no initial sur� gery were evaluated by the same methods as those used for the e�perimental knees. The mean value for the 2 knees of each animal was used for calculations. 3 animals were killed at the age of 22 weeks, 3 at age 34 weeks (22 + 12), 2 at 46 weeks (22 + 24), and 2 at 58 weeks (22 + 36). The intention was to add these knees to a larger number of control specimens, harvesting the patella as a control specimen from knees with condylar defects and vice versa. However, since surgery had been performed on another joint surface within the same knee, the latter control specimens were e�cluded. Thus, due to the low number of "control" specimens, the data e�tracted were Figure 1. The experimental animals gained weight throughout the experimental period similar to that of the control animals. The number of experimental animals at time zero (age 22 weeks) equalled the total number of animals evaluated at the follow-ups (37), since they were all measured preoperatively. The numbers of experimental animals at 12, 24, and 36 weeks were 8, 11, and 18, respectively. The numbers of animals without surgery at time 0, 12, 24, and 36 weeks were 3, 3, 2, and 2, respectively.

Surgical technique
Through a medial parapatellar incision, a defect with a diam� eter of 4 mm was created in the patella of one knee (random� ized to left or right) and in the medial femoral condyle of the contralateral knee, as previously described (Aroen et al. 2005) ( Figure 2). A stereo microscope and small instruments were used (Figure 3), and care was taken to avoid any damage to the rims of the defects or to the underlying calcified cartilage. The defects were left untreated, the joint was irrigated, hemosta� sis was achieved, and wound closure was performed as previ� ously described (Aroen et al. 2005).

Killing of animals, macroscopic evaluation, and preparation for histological analysis
The animals were killed and synovial fluid was obtained as previously described (Aroen et al. 2005). The patellas and the femoral condyles were dissected free and gross morphological grading was performed. Changes to the cartilage correspond� ing to �CRS grade 1-2 and/or small osteophytes were catego� rized as minor changes, whereas changes e�ceeding that were categorized as major. The specimens were immersed in phos� phate�buffered 4% paraformaldehyde for 1 week and decalci� fied in 20% formic acid until the bone was soft enough for sec� tioning. Cubes, measuring 8 × 8 mm and containing the defect at one side, were harvested from the specimens, dehydrated in graded alcohol and embedded in an epo�y resin. The patel� las and condyles with no defects were handled in the same manner as the e�perimental specimens. To define the region for histological analysis, an area corresponding to the location of the e�perimental defects was outlined using the same kind of 4�mm biopsy punch as used when creating the defects in the e�perimental knees.

Histology
The cubes were sectioned from one longitudinal surface, with random orientation of the defect (anterior�posterior or medial� lateral) (Gundersen et al. 1988). From the point at which the rim of the defect was reached, sections-each 1-2 µm thickwere captured at 5 levels, each level 700 µm further into the defect. This technique ensured that one of the sections would represent a level at a ma�imum distance of 350 µm from the very center of the defect ( Figure 4). The sections were stained with tolouidin blue and photographed at 40× magnification using a digital camera mounted on the microscope. An inter� active semiautomatic image analysis program (Analysis Pro; Olympus Soft �maging Solutions, Münster, Germany) was used for all measurements. The section closest to the center of each defect (largest diameter) was included in the statisti� cal analyses. The observer was blind as to the location of the defects.

Estimation of tissue filling
The borders of the defects were identified by the interfaces between presumed original cartilage and newly formed fibro� cartilaginous/fibrous repair tissue on both sides. The midpoint along the tidemark of the defect was defined. From the mid�  point, sectors 0.5 mm in length were marked along the tide� mark to each side until 2 mm from midpoint was reached on each side, representing the tidemark at the base of the "shoul� ders" of the 4�mm defect. Cartilage height was measured at these 2 shoulder points, as was tissue height at the 7 inter� section points of the 0.5�mm sectors ( Figure 5). Tissue filling was estimated by relating the mean height of tissue at the 7 defined points to the mean cartilage height of the 2 shoulders of the same defect, e�pressed as percentage filling. Whenever one of the shoulders was not measurable technically, the one shoulder that was left served as the reference. This occurred in sections from 1 patellar defect and 1 condylar defect.

Estimation of the density of mineralized subchondral tissue
The density of the mineralized subchondral tissue was esti� mated in a region immediately beneath the condral defect. The depth of the region was 1.5 mm measured from the tidemark, thus the shape of the region depended on the curvature of the tidemark ( Figure 5). The morphometry was performed by point counting (Gundersen et al. 1988) using computer soft� using computer soft� using computer soft� ware. A grid with 0.3 mm between test lines was superimposed on the micrograph. The intersections between the test lines served as test points. Subchondral mineralized tissue density was e�pressed as the number of test points overlying mineral� ized tissue relative to the total number of test points within the region of interest. The counting was repeated twice for each defect with the orientations of the grid randomly selected on both occasions. The mean value of the 2 measurements was used for statistical analysis. A similar technique was used by Løken et al. (2008), demonstrating a variance of less than 10% between measurements and between observer.
Analysis of synovial fluid 41 animals had initial surgery at 22 weeks of age. From 60 of the knees, a wash�out sample of the synovial fluid containing a minimum of 0.75 mL collected before surgery (time zero) was available for analysis of proteoglycan content using standard EL�SA methodology (Messner et al. 1993b). A second wash� out sample, collected at killing, was available for analysis from 69 of the knees. The discrepancy was mainly due to dry taps. 58 knees had samples available for analysis both at time zero and at killing, whereas 24 animals had samples available from both knees at both time zero and at killing. The samples collected from the e�perimental animals were all analyzed as one batch.

Statistics
Based on previous e�perimental studies ( Aroen et al. 2006), a filling difference of more than 25% was considered to be a proper level to discharge the H 01 �hypothesis of no difference in tissue filling between the patellas and the condyles. Since the e�perimental animals all underwent surgery in both knees, with patellar defects in one and defects of the MFC in the other, they would serve as their own control regarding tissue filling of the defects, and thus a paired Student t�test could be applied. Pre�e�perimental analysis to detect sample size using a power of 0.80 and a significance level of 0.05 and a standard deviation for the differences of less than 24% indicated a need for 9 animals in each group for this purpose. To evaluate the difference in tissue filling from one follow�up time point to another regarding each location separately, the need for ani� mals was estimated to be 12 for each of the 3 follow�up time points-due to an unpaired e�perimental situation. Power and sample size estimations were not performed for the second� ary endpoints. To cover up for possible loss of animals during follow�up, 41 rabbits were to be included in the study. The difference in tissue filling between patellar defects and MFC defects was calculated and evaluated by one�way A�OVA and paired Student t�tests. Tissue filling with time was evaluated by one�way A�OVA and unpaired t�tests. Dif� ferences in mineralized tissue density between patellas and condyles-and the changes with time-were analyzed by one�way A�OVA and paired and unpaired t�tests. The change in proteoglycan content of synovial fluid from time zero to follow�up, named delta (Δ), was evaluated by A�OVA and paired t�tests. �nteractions between time and location were investigated; dependent observations at the individual level were accounted for by pairwise computation of the difference between the deltas of the knees, the results being used in an A�OVA model with time to follow�up as group factor. The rise in proteoglycan content (delta) for each location at each follow�up was further analyzed by paired t�tests, the level of significance being corrected according to Bonferroni. The SPSS software package version 14 was used for statistical analysis.

Results
Complications �one of the animals died during surgery. Postoperatively, visual observation did not reveal any harmful effects on gait in the animals and no differences in level of activity or motion pattern between legs were observed. 5 of the 41 e�perimental animals died une�pectedly of unknown causes during follow� up. The knees were all unaffected, the deaths were all within 2 weeks of a follow�up time point, and the specimens were therefore kept in the study. Additionally, 4 animals had to be killed during follow�up due to impaired general health. One of these was killed 2 weeks after initial surgery and was e�cluded from the study. Of all the animals, complications related to the knee were observed in 14 cases: 7 knees were e�cluded due to patellar dislocation, 1 knee because of infection, and 6 knees due to technical failures in histology preparation.

Macroscopic changes
Of the knees with defects but no complications, none of them had major degenerative changes at any follow�up point. Some minor degenerative changes were observed in 8 of 69 e�peri� mental knees (Table 1). There were no statistically significant differences in the number of knees with minor degenerative changes that were not related to either the location of the defect or time. Furthermore, there was no correlation between either the changes in subchondral mineralized tissue density (p = 1.0) or the changes in synovial fluid proteoglycan content (p = 0.6) and the minor degenerative changes observed.

Filling of the chondral defects
There was a significantly higher degree of tissue filling in patellar defects compared to condylar defects at all follow�ups ( Table 2). The difference in tissue filling between the 2 loca� tions remained similar with time (p = 0.2). Both the patellar and condylar defects showed a change in filling with time (p = 0.02 and p = 0.02, respectively) ( Figure 6), the change being apparent at the interval from 24 to 36 weeks (p = 0.003 for patellar defects and p < 0.001 for condylar defects).

Mineralized subchondral tissue
The mineralized subchondral tissue density of the patel� las with defects decreased with time (p = 0.01), whereas the density of condyles with defects remained similar with time (p = 0.9). The reduction in patellar density mainly occurred   in the interval between 12 and 24 weeks (p = 0.01), contribut� ing to a major reduction in density between 12 and 36 weeks (p < 0.001) (Figure 7). Although the descriptive statistics (95% C�) and one�way A�OVA applied on the computed pairwise differences between patellar and condylar defects showed a higher density in patellas with defects than in condyles with defects at all time points, the difference was reduced in the interval from 24 to 36 weeks (p = 0.02). These findings were supported by paired Student t�tests comparing patellar defects with condylar defects, the difference being statistically highly significant at 12 and 24 weeks, whereas the significance of the difference was borderline at 36 weeks-and not significant if Bonferroni's correction was used.
Proteoglycan content of synovial fluid A higher proteoglycan content was detected at 12 weeks than at time zero both for knees with patellar defects (p = 0.007) and for those with condylar defects (p = 0.003), whereas the values at 24 and 36 weeks were similar to those at time zero (Figure 8). There was no interaction between defect location and time to follow�up, and there was no effect of location on the change in proteoglycan content.

Effect of intraarticular location on defect filling
We found a higher degree of filling in the patellar defects than in the medial femur condylar defects at the 24� and 36�week follow�ups. At 36 weeks, the mean difference in filling was 28% with a standard deviation of 27% (p = 0.001). The large variance in spontaneous filling of �CRS grade 3b defects has been emphasized by other authors (Breinan et al. 1997). Tissue filling is to some e�tent assumed to be crucial in car� tilage restoration of a chondral defect. However, the critical amount of tissue filling necessary to discriminate one clinical outcome from another regarding joint function, pain, disabil� ity, and reduced risk of osteoarthritis, is not well understood. According to our power and sample size estimation, a filling difference of more than 25% was considered to be a proper level to discard the H 01 �hypothesis of no difference in tissue filling between locations in the patella and the medial femoral condyle. The mean differences at 12 and 24 weeks were both below 25%. At 36 weeks, however, the mean difference was 28%; thus, the H 01 �hypothesis could be discarded. The finding indicates that the intraarticular location influences the natural history of filling of a pure chondral defect �CRS grade 3b in a long�term follow�up. To our knowledge, this issue has not been thoroughly investigated previously. The potential effect of the intraarticular location on the outcome of surgical repair of chondral defects (�CRS grade 3b) has been discussed in previous papers (Breinan et al. 1997(Breinan et al. , 2000. Breinan et al.
(1997) compared autologous chondrocyte implantation (AC�) in trochlear defects to untreated controls in dogs. They did not detect any significant effect of the AC� on the amount of defect filling, in contrast to Brittberg et al. (1996) who did the same comparison of AC� and controls in patellar defects in rabbits. Breinan et al. (1997) suggested the intraarticular location of the defect (trochlea vs. patella) to be one possi� ble e�planation for the discrepancy in results. �n a later study using the same dog model with 2 chondral defects (�CRS grade 3b) in trochlea, this group found a difference between pro�imal and distal defects in the percentage of reparative tissue that was fibrocartilage (p = 0.02), suggesting that the intraarticular location may play a role in cartilage restoration (Breinan et al. 2000). A discrepancy in results after cartilage repair, possibly related to the intraarticular location, has been noted in clinical studies also, the femoral condyle being the   most favorable location (Brittberg et al. 1994, Hangody and Fules 2003, Krishnan et al. 2006.
We found an effect of the intraarticular location to be a predictive factor for the outcome of natural history of tissue filling of an �CRS grade 3b defect in the rabbit knee. We believe that knowledge of intrinsic factors that influence the outcome is essential in evaluating different aspects of carti� lage restoration.

Effect of time on defect filling
We observed a change in tissue filling of both the patellar and condylar defects with time. This difference remained similar with time, indicating that the changes in tissue filling were similar at the two locations. At both locations, the amount of filling increased from 24 to 36 weeks. This observation shows that time to follow�up is an important parameter in evaluat� ing the results of natural history tissue filling in e�perimental models. The effect of time on cartilage restoration has been well studied in animal models involving defects correspond� ing to �CRS grade 4 (Shapiro et al. 1993, Lietman et al. 2002. On the other hand, the natural time course of �CRS grade 3 defects is less well known. �n their untreated control defects, Grande et al. (1989) observed 17% spontaneous filling of a 3�mm "full�thickness" defect (through all chondral layers into the calcified zone, corresponding to �CRS grade 3c) in the rabbit patella at 6 weeks of follow�up. Using the same rabbit model as Grande, Brittberg et al. (1996) increased the follow�up time and reported 29% spontaneous filling at 12 weeks, which was similar to the 32% filling of patellar defects obtained at 12 weeks in our study. However, in our study the amount of tissue filling increased to 49% at the 36�week follow�up. The increase in spontaneous tissue fill� ing after 12 weeks is in line with studies using other animal models. Breinan et al. (1998) reported 35% spontaneous fill� ing of a 4�mm defect corresponding to �CRS grade 3b in the knee trochlea of dogs at 3 months. �n a separate study using the same dog model, these authors obtained 41% spontaneous filling at 12 months, which increased to 76% at 18 months (Breinan et al. 1997). �n our study, for both locations the amount of filling tended to decline from 12 to 24 weeks and then increased from 24 to 36 weeks. This observation is probably not in conflict with those of Frisbie et al. (1999) investigating 1 cm 2 "full thickness" defects (removing all the calcified cartilage, but preserving the bone plate; corresponding to �CRS grade 3c) in a horse model. Due to merging of the data from the groups in various ways, the results presented are not easy to interpret, but the mean filling of the treated defects and the control defects together was 44% at 4 months and 54% at 12 months; the filling of the controls when the 2 time points were merged together was lower than the filling of the treated defects. �n another study using the same model, the same authors reported 52% filling of control defects at 8 weeks of follow�up (Frisbie et al. 2003). The combined results from these 2 studies seem to give the same natural history for filling of a defect as we found: there was a high percentage of filling at an early follow�up, with a tendency towards decrease at a medium�time follow�up, and increasing again at long�term follow�up. We observed such a time course for both locations, but we cannot e�plain the phe� nomenon. We believe, however, that knowledge of the natural time course in a given model is essential for evaluation of dif� ferent aspects of cartilage restoration.
Mineralized subchondral tissue density �n contrast to the density of the subchondral tissue in the con� dyles, the density of the mineralized subchondral tissue in the patellas changed with time. This indicates that the intraarticu� lar location is also a predictive factor for subchondral bone changes related to chondral defects of �CRS grade 3b in the rabbit knee. To our knowledge, this issue has not been dis� cussed in any of the previous literature.
One weakness of our study was the lack of sufficient numbers of control patellas and condyles for evaluation of subchondral bone changes. Our intention was to use patellas from knees with condylar defects, and femoral condyles from knees with patellar defects as controls. Additional animals were added only to make up a sufficient number of control specimens. We learned, however, that there is a potential effect of chronic cartilage lesions in one articular surface to cause changes to the subchondral bone of other articular surfaces within the same joint (Marijnissen et al. 2002, Sniekers et al. 2008. The contralateral patella/condyle of the e�perimental animals was therefore e�cluded, leaving us with a number of "controls" that was too small for statistical analysis. However, the observation that the density of mineralized tissue in patel� las changed with time-whereas in condyles it did not-may be an important finding. Subchondral bone changes are associ� ated with the initiation and progression of osteoarthritis (OA) (Radin and Rose 1986, Burr 1998, 2004. There are, however, no objective criteria regarding the degree of changes that would indicate a substantial initiation or progression of OA. Thus, no estimations of power or sample size were performed considering the changes in the density of subchondral miner� alized tissue. Whether the statistically significant changes we found have any clinically implications thus remains uncertain. Some authors have emphasized that changes in the calcified cartilage, being part of the mineralized subchondral tissue, may play a role in the pathogenesis of OA (Burr 2004). Thus, defects involving iatrogen damage to the calcified cartilage (�CRS grade 3c) should probably be distinguished from pure chondral lesions (�CRS grade 3b) in evaluating the natural his� tory of the subchondral mineralized tissue. The impairment of a well�defined pure chondral defect (�CRS grade 3b) on the subchondral bone is less well known. By merging data from 12 and 18 months of follow�up in dogs, Breinan et al. (1997) reported resorption of the subchondral bone leading to mod� erate to severe bone loss in 3 of 14 untreated �CRS grade 3b defects in the trochlea. They offered no e�planation for the changes, but raised the possibility of having caused damage to the calcified cartilage during surgery, made evident by study� ing fresh defects in cadaver dogs. To reduce the risk of damage to the tidemark and the underlying calcified cartilage in our study, we used a stereo microscope and small instruments in preparing the defect.
We found a reduction in the density of mineralized subchon� dral tissue from 12 to 36 weeks in the patellas with defects. The time course is in agreement with that of changes in cor� responding parameters reported in e�perimental OA models. �n their model of OA in dogs, Sniekers et al. (2008) reported a 6% increase in bone volume fraction at ten weeks, which sub� sided to a 13% loss at 20 weeks. The finding of reduced bone volume fraction combined with thinning of the subchondral bone plate as well in early degenerative joint disease is also in agreement with other studies (Dedrick et al. 1993).
Although the filling of the defects in the condyles was lower at all time points, the mineralized subchondral tissue in the condyles seemed to tolerate the defect better, showing no detectable changes in density with time. Thus, the intraarticu� lar location may be a predictive factor for subchondral bone changes related to chondral defects of �CRS grade 3b in the rabbit knee.

Proteoglycan content of synovial fluid
A rise in synovial fluid proteoglycan content was detected in the knees of animals killed 12 weeks after surgery. There was no statistically significant difference between knees with patellar defects and knees with condylar defects. The rise observed at 12 weeks subsided with time, being similar to time�zero values at 24 and 36 weeks of follow�up.
�ncreased concentrations of proteoglycan fragments in the joint fluid have been found to be associated with trauma and surgery in humans (Lohmander et al. 1989, Odenbring et al. 1991, and with increasing knee OA in rabbits (Messner et al. 1993a). �n their rabbit OA model, Messner et al. found an increase at 3 months which they e�plained by the effect of surgery, the values decreasing to normal at 6 months and then increasing again at 12 months, which they e�plained as pos� sibly being related to initial degeneration. These findings are not in conflict with the observations in our study; there was a rise in proteoglycan content at 12 weeks, subsiding to values similar to time zero at the 24� and 36�week follow�ups. Con� sequently, the proteoglycan concentrations of synovial fluid in our study did not indicate a degenerative process during the observation period.

Weaknesses of the study
The sample size estimation indicated a need for 12 animals for each of the 3 follow�ups. A sufficient number was obtained at 36 weeks of follow�up, whereas the numbers at 12 and 24 weeks were below that due to a high number of complications and failures in histological preparation. The data obtained at 36 weeks thus represent the most valuable information. Com� plications and loss of animals related to the rabbit model have also been described by other authors (Brittberg et al. 1996).

Strengths of the study
The importance of defining the depth of e�perimentally pro� duced defects in relation to the different layers of the joint organ-and evaluation of the results in relation to that-has been emphasized by others (Brittberg et al. 1996, Breinan et al. 1997, Frisbie et al. 1999, Hunziker 1999, Burr 2004. Even the bias of not removing all the tissue of a layer as intended, or causing damage to the layer beneath, has been a topic in discussing the results of cartilage restoration (Breinan et al. 1997, Frisbie et al. 1999). Methodologically, our study was strengthened by the use of a stereo microscope, which ensured removal of all the cartilage without causing damage to the calcified layer and the tissue beneath in preparing the defects. �n addition, we used the most commonly applied animal model in cartilage repair.