Article Text

Extended report
Histopathological changes in the human posterior cruciate ligament during aging and osteoarthritis: correlations with anterior cruciate ligament and cartilage changes
  1. Yadin D Levy1,2,
  2. Akihiko Hasegawa2,
  3. Shantanu Patil1,
  4. James A Koziol2,
  5. Martin K Lotz2,
  6. Darryl D D'Lima1,2
  1. 1Shiley Center for Orthopaedic Research and Education at Scripps Clinic, La Jolla, California, USA
  2. 2Department of Molecular and Experimental Medicine, The Scripps Research Institute, La Jolla, California, USA
  1. Correspondence to Professor Martin K Lotz, Department of Molecular and Experimental Medicine, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA; mlotz{at}


Objectives To determine the histological patterns of posterior cruciate ligament (PCL) degeneration during aging and in relation to changes in articular cartilage and anterior cruciate ligament (ACL) across the entire adult age spectrum.

Methods Human knee joints (n=120 from 65 donors) were processed within 72 h of postmortem. Articular cartilage surfaces were graded macroscopically. Each PCL was histologically evaluated for inflammation, mucinous changes, chondroid metaplasia, cystic changes and orientation of collagen fibres. The severity of PCL degeneration was classified as normal, mild, moderate or severe. PCL scores were compared to ACL and cartilage scores from the same knees.

Results All knees had intact PCL. Histologically, 6% were normal, 76% showed mild, 12% moderate and 9% severe degeneration. Fibre disorientation was the most prevalent and severe change. Histological grades of PCL and ACL correlated, but significantly fewer PCL than ACL showed severe changes. There was a weaker correlation between aging and total histological PCL scores (R=0.26) compared to aging and ACL scores (R=0.42). ACL scores correlated with cartilage scores (R=0.54) while PCL scores increased with the severity of osteoarthritis from grades 0 to III but not between osteoarthritis grades III–IV (R=0.32). In knees with ruptured ACL, the PCL scores correlated with cartilage scores of the lateral compartment.

Conclusions PCL histopathological changes were less severe than in the ACL. PCL degeneration was associated with ACL and cartilage damage. The lack of correlation with age indicates independent pathways for PCL versus ACL degeneration.

  • Osteoarthritis
  • Knee Osteoarthritis
  • Inflammation

Statistics from


Osteoarthritis, the most common joint disorder,1 has a multifactorial aetiology with systemic factors that predispose to the disease and local factors that determine its distribution and severity.2 Aging and trauma are two of the strongest risk factors for osteoarthritis.2 It is not yet established if the osteoarthritis process begins in a specific joint structure, but it is known that in later stages it manifests in all joint tissues.3–6

The posterior cruciate ligament (PCL) is an important knee structure. Together with the anterior cruciate ligament (ACL) they are essential for knee kinematics in anteroposterior tibial translation, tibial rotation, and as secondary restraints to valgus/varus forces.7–9 Rupture of the PCL is most commonly associated with severe knee trauma and multiligament injury10 but rarely occurs spontaneously or in isolation. This is in contrast to degeneration of the ACL, which has been reported to occur in up to 25% of patients with severe knee osteoarthritis.4 ,11 Besides the importance of the PCL in knee biomechanics, it is an important knee structure in patients with end-stage osteoarthritis undergoing total knee arthroplasty (TKA). PCL retaining TKA designs rely on the integrity of the PCL.12

Histological changes in the PCL are observed in knees undergoing TKA for severe osteoarthritis.11 These changes include mucoid degeneration, disorganisation of the collagen fibres and cystic changes.11 ,13 Moreover, during TKA large numbers of knees without previous known trauma were diagnosed with ruptured or absent ACL, whereas the PCL was macroscopically intact.11 These knees are characterised by an increased severity of medial and lateral compartment osteoarthritis.14 The concomitant histopathological changes in the PCL and their interplay with changes in the ACL and cartilage are not well understood. Even less is known about intrisic age-related changes in the PCL and their relationship with ACL and cartilage at early stages of the disease process.

The objectives of this study were to identify patterns and severity of a comprehensive set of histological changes in the PCL from a large number of human knee joints across the entire adult age range and at all stages of cartilage degradation in donors with no history of previous joint trauma and to determine the correlation of changes in the PCL with the ACL and the articular cartilage.

Materials and methods

Procurement of human knee joints

Human knee joints were obtained from tissue banks with approval of the Scripps Human Subjects Committee. Joints were processed within 72 h of postmortem. In this study, 120 human knee joints were analysed from 30 male and 35 female donors, with a mean age of 66.7±19 years (range 23–92 years) and mean body mass index (BMI) 24.8±7 (range 12.5–43.8). The ethnic distribution of the donors was as follows: Hispanic (n=26), Caucasian (n=80), black (n=2) and unknown (n=12). From 55 of 65 donors we obtained both knees. Subjects with a history of knee trauma or surgery were excluded.

Ligament analysis

Evaluation of the ligaments was performed as described previously.11 ,15–17 The PCL and the ACL were resected at the insertion sites on the femur and tibia. The macroscopic appearance of the ligaments was classified in three stages: normal, abnormal (thinner than normal and sclerotic), or ruptured.15 For histology, the samples were immediately fixed in Z-Fix (Anatech, Battle Creek, Michigan, USA). From each ACL and PCL two cuts were prepared and used for histological evaluation, a transverse cut at the proximal one-third of the ligament and a longitudinal cut through the centre of the ligament. The samples were from the proximal one-third of the ACL substance and femur attachment site where ACL tears frequently occur.18 ,19 We used only the ligament region and not the non-mineralised interface, mineralised interface, or bone for histological analysis. Samples were embedded in paraffin. Four micrometer-thick sections were cut and stained with H&E.

The PCL and ACL sections were graded histologically using a modification of previous scoring systems.11 The following categories were examined and scored for each ligament: (1) inflammation in the ligament substance; (2) mucoid degeneration; (3) chondroid metaplasia; (4) cystic changes; and (5) orientation of collagen fibres. Each category was scored on both transverse and longitudinal sections and a single score was assigned for each ligament. The histological changes were scored and graded as follows: 0, no changes; 0.5, minimal changes; 1, mild changes; 2, moderate changes; 3, severe changes. The highest summed score of ligament degeneration (total PCL or ACL score) was 15 if all five histological categories were scored severe.20

The PCL were graded by two independent readers to assess intra-observer and inter-observer reliability. At the time of sample evaluation the readers were blinded and unaware of the age, sex and cartilage or ACL scores of the same knees. Each reader performed two readings of the entire sample set. The intraclass correlation of the first reader was 0.85, and 0.91 for the second reader. The interclass correlation between readers was 0.87.

Morphological analysis of articular cartilage

All cartilage surfaces (femoral condyles, trochlea and tibial plateaus) were graded macroscopically using a modified Outerbridge scoring system.21 We established a more detailed scoring method based on the International Cartilage Repair Society knee map (see supplementary figure S1, available online only).22 The articular cartilage of the knee was divided into several compartments: three areas in the trochlea and nine areas each in each femoral condyle and tibial plateau. Each of these 39 areas was scored individually, and total compartment and total knee scores were calculated. For each area a score ranging from 1 to 4 was assigned as follows: 1, intact surface; 2, fibrillation; 3, fissuring; 4, erosion. The total knee cartilage scores ranged from 39 (normal) to 156 (maximum severity). For statistical analyses, the total knee cartilage scores were translated into grades 0–IV (grade 0, normal when total score was 39; grade I, minimal change when total score was 40–58; grade II, mild change when total score was 59–78; grade III, moderate change when total score was 79–97; and grade IV, severe change when total score was higher than 98). Donor distribution among the different cartilage grades was as follows: grade 0 (n=9), grade I (n=51), grade II (n=31), grade III (n =18) and grade IV (n=11).

Statistical analysis

Summary statistics are reported as mean±SD. Spearman's non-parametric correlation coefficients were used to assess strengths of bivariate associations. 95% CI for the reported Spearman correlation coefficients were obtained from the bootstrap percentile method, based on 1000 samples. Levels of the CI are unadjusted, as our emphasis is on estimation of the strength of bivariate associations rather than inference. We used general linear models (GLM) to investigate associations between total PCL scores and the severity of macroscopic changes, and between total PCL and ACL scores. Similarly, we used GLM to examine relationships between PCL and ACL histopathology and total cartilage scores, with age, gender and BMI. In these latter analyses, gender was dichotomous, and age and BMI were continuous variables. Stratified paired t tests were also used to compare component and total PCL and ACL scores; the stratification factor was either cartilage grade (0–IV) or separately age, categorised into quartiles. These analyses simultaneously allow the assessment of how mean scores differ across the cartilage grades, or age quartiles. Two-sided p values of 0.05 or less were considered indicative of statistical significance. We utilised a non-parmetric technique, LOESS smoothing,23 to depict the relationship between PCL, ACL and cartilage scores with age. Here, we fitted first degree LOESS curves, with tricube kernel and 25 grid points. Calculations were performed using STATA V.9.2 and SPSS V.16.0.


Relationship between PCL and ACL histopathology and age, gender and BMI

On macroscopic evaluation all PCL in the 120 knees were intact. The mean total PCL and ACL histological scores were 3.82±2.3 and 4.02±2.7, respectively. The total PCL histological grades were as follows: 3% (n=4) categorised as normal; 76% (n=91) mild; 12% (n=14) moderate and 9% (n=11) severe. The total ACL histological grades were as follows: 3% (n=4) normal; 69% (n=83) mild; 15% (n=18) moderate; and 13% (n=15) severe. Total PCL histological scores significantly increased with the severity of macroscopic ACL changes (normal, 3.39±1.89; n=68; abnormal, 4.24±2.59; n=40; ruptured; 4.83±2.72, n=12; F2,117=3.21, p=0.044).

From multivariable GLM analyses, the total PCL histopathological scores differed significantly by gender and age, but not BMI (see supplementary table S1A, available online only). In particular, mean PCL scores for women were significantly lower than mean PCL scores for men (women, 3.45±2.15; men, 4.28±2.18). In comparison, the total ACL histopathological scores, as well as the total cartilage scores, differed significantly by age, but not by gender or BMI (see supplementary table S1B,C, available online only). The trends of increasing total PCL, ACL and cartilage scores with age are depicted in figure 1. Total cartilage score shows the strongest correlation with age (Spearman's r=0.75, 95% CI 0.66 to 0.82), followed by total ACL score (r=0.42, 95% CI 0.23 to 0.58) and then total PCL score (r=0.26, 95% CI 0.08 to 0.42). The highest scores were seen in the 80–89-year-old group for both ligaments (PCL 5±3; ACL 6.1±3.7). The total PCL score was consistently higher in each decade until 60 years. This trend reversed after the age of 60 years when the mean total ACL score continued to increase more than the PCL score.

Figure 1

Bivariate scatterplots of total posterior cruciate ligament (PCL), total anterior cruciate ligament (ACL), and total cartilage scores, and associated ages. Also depicted are the LOESS regressions (smoothers), indicative of increasing scores with age. Spearman's correlations r between age and total PCL (left panel), total ACL (middle panel) and total cartilage scores (right panel) are 0.26 (95% CI 0.08 to 0.42), 0.42 (95% CI 0.23 to 0.58) and 0.75 (95% CI 0.66 to 0.82), respectively. This figure is only reproduced in colour in the online version.

PCL, ACL and cartilage changes

Total knee cartilage scores were more strongly correlated with total ACL scores (Spearman's r=0.54, 95% CI 0.39 to 0.66) than with total PCL scores (r=0.32, 95% CI 0.13 to 0.49) (figure 2). Both ACL scores and PCL scores differed significantly by grade (severity of cartilage changes): ACL, F4, 112=13.02, p<0.001; PCL, F4, 115=4.67, p=0.002. PCL scores increased steadily with the severity of cartilage changes from grade 0 to grade III, but there was no further increase in grade IV, whereas ACL scores increased with each cartilage grade (figure 2A). Components of ACL and PCL scores also tended to increase with increasing osteoarthritis grade (figure 2B–F): inflammation, F4,111=2.63, p=0.038; mucoid, F4,111=4.34, p=0.027; chondroid, F4,111=2.54, p=0.044; cystic changes, F4,111=4.36, p=0.0026; fibre orientation, F4,111=11.10, p<0.001; total scores, F4,111=11.42, p<0.001. ACL but not PCL cystic change scores increased significantly at grade IV compared to the less severe grades, and ACL fibre orientation scores increased more rapidly than the corresponding PCL scores as cartilage grade severity increased.

Figure 2

Total and specific posterior cruciate ligament (PCL) and anterior cruciate ligament (ACL) scores versus cartilage grade. Values depicted are the mean ligament scores at each specific cartilage grade, separately for PCL and ACL. Total PCL and ACL histological score (A), inflammation (B), mucoid changes (C), chondroid changes (D), cystic changes (E) and fibre orientation (F). Mean total and component ACL and PCL scores tended to increase with increasing osteoarthritis (OA) grade: total scores, F4,111=11.42, p<0.001; inflammation, F4,111=2.63, p=0.038; mucoid, F4,111=4.34, p=0.027; chondroid, F4,111=2.54, p=0.044; cystic changes, F4,111=4.36, p=0.0026; fibre orientation, F4,111=11.10, p<0.001. This figure is only reproduced in colour in the online version.

Correlation of PCL changes with knee compartmental cartilage scores

We determined if ligament degeneration was linked to local cartilage degeneration in specific knee compartments. PCL scores were weakly correlated with compartmental cartilage scores (medial femoral condyle r=0.27, 95% CI 0.07 to 0.45; medial tibial plateau r=0.30, 95% CI 0.11 to 0.47; lateral femoral condyle r=0.24, 95% CI 0.06 to 0.41; lateral tibial plateau r=0.29, 95% CI 0.09 to 0.46; trochlea r=0.28, 95% CI 0.09 to 0.45) (see supplementary figure S2, available online only). In contrast, ACL scores were more highly correlated with compartmental scores, specifically the medial compartment (medial femoral condyle r=0.5, 95% CI 0.38 to 0.65; medial tibial plateau r=0.46, 95% CI 0.30 to 0.60; lateral femoral condyle r=0.48, 95% CI 0.33 to 0.60; lateral tibial plateau r=0.44, 95% CI 0.27 to 0.58; trochlea r=0.43, 95% CI 0.26 to 0.58).

Specific histological changes in the PCL and the ACL

When the ACL and PCL ligament scores were compared in a pairwise fashion, ACL inflammation scores were found to be significantly higher than the corresponding PCL scores (t116=3.63, p<0.001), and PCL chondroid metaplasia scores were significantly greater than their ACL counterparts (t115=−2.71, p=0.009). No significant pairwise differences between ACL and PCL scores were found with inflammation, mucoid degeneration, fibre organisation, or total scores. In figure 3, the mean ACL and PCL ligament scores are depicted, stratified by age quartiles. Mucoid scores, cystic changes scores, orientation scores, as well as total scores significantly increase with age (mucoid, F3,112=4.13, p=0.008; cystic changes, F3,112=3.82, p=0.012; orientation, F3,112=11.72, p<0.0001; total, F3,112=8.20, p<0.0001); inflammation scores and chondroid scores tended not to differ significantly across the age quartiles (inflammation, F3,112=1.45, p=0.23; chondroid, F3,112=1.01, p=0.39).

Figure 3

Total and specific posterior cruciate ligament (PCL) and anterior cruciate ligament (ACL) scores versus age. Ages were divided into four quartiles; within each quartile, means and associated 95% CI of total and specific histological scores are shown, separately for PCL (red) and ACL (blue). Total PCL and ACL histological score (A), inflammation (B), mucoid changes (C), chondroid changes (D), cystic changes (E) and fibre orientation (F). In paired analyses, ACL inflammation scores were found to be significantly higher than the corresponding PCL scores (t116=3.63, p<0.001), and PCL chondroid metaplasia scores were significantly greater than their ACL counterparts (t115=−2.71, p=0.009). No significant pairwise differences between ACL and PCL scores were found with inflammation, mucoid degeneration, fibre organisation, or total scores. This figure is only reproduced in colour in the online version.


Among the different types of histopathological changes, PCL inflammation was the least apparent, with an average score of 0.46±0.7 and only nine knees (8%) demonstrated moderate to severe inflammation. By contrast, the average ACL inflammation score was 0.82±1 and 30 knees (25%) demonstrated moderate to severe inflammatory changes. The PCL inflammation score did not significantly correlate with any of the other PCL histological criteria, with the exception of PCL orientation (r=0.33, 95% CI 0.18 to 0.48), donor age, total cartilage score, or ACL inflammation score. The PCL scores were lower than the ACL scores in all osteoarthritis groups, especially in osteoarthritis grade II (mean PCL 0.3±0.4; mean ACL 1.1±1.1) (figure 2B).

Mucoid degeneration

Mucoid degeneration was observed in 71% of the PCL and correlated strongly with PCL chondroid metaplasia (r=0.73, 95% CI 0.59 to 0.83). When compared to the ACL mucoid score across cartilage grades, the PCL mucoid degeneration score was higher than the ACL scores in cartilage grades 0–III (figure 2C).

Chondroid metaplasia

Chondroid metaplasia was observed in 75% of the PCL and correlated well with PCL mucoid degeneration (r=0.73, 95% CI 0.59 to 0.83). With respect to age, chondroid metaplasia scores increased continuously up to the 6th decade, followed thereafter by slightly decreased and constant scores. The average chondroid metaplasia scores for the PCL were highest for osteoarthritis grades II and III. On the other hand, ACL chondroid metaplasia scores increased consistently with cartilage damage. PCL chondroid metaplasia scores were higher than in ACL for cartilage grades 0–III but less in cartilage grade IV (figure 2D).

Cystic changes

Cystic changes were observed in 38% of PCL. The PCL cystic score did not significantly correlate with age, other PCL histological changes, total cartilage score, or total ACL score. The PCL cystic score was equal or slightly higher for cartilage grades 0–III compared to the ACL scores. For cartilage grade IV the ACL scored more (mean 1.7±1.1) than the PCL (mean 0.5±0.7). Again, the ACL score for cartilage grade IV was higher than the PCL score (figure 2E).

Collagen fibre organisation

Disorganisation of collagen fibres was observed in 94% of PCL. The average PCL fibre disorganisation score was the highest among the different histopathological criteria (mean 1.2±0.7). The PCL fibre disorganisation score was significantly correlated with age (r=0.24, 95% CI 0.08 to 0.40), knee cartilage score (r=0.29, 95% CI 0.10 to 0.47), PCL inflammation score (r=0.33, 95% CI 0.18 to 0.48) and PCL mucoid score (r=0.20, 95% CI 0.01 to 0.39). With respect to age, the PCL scores were higher than the ACL up to the 6th decade, and thereafter the ACL scores were higher. PCL fibre disorganisation scores increased with osteoarthritis severity and were higher than the ACL for osteoarthritis grade 0, but thereafter the ACL scores were higher than PCL scores. Among the different criteria for PCL evaluation, fibre disorientation was the only parameter to increase continuously with increasing cartilage grades (figure 2F).

Aging-related changes in PCL, ACL and cartilage

To identify the earliest changes in the PCL we selected nine knees from donors younger than 45 years with normal cartilage (grade 0). In this subset the PCL histological scores were compared to the ACL scores (figure 4). The first changes observed in the PCL were fibre disorganisation, mucoid and chondroid changes.

Figure 4

Total and specific posterior cruciate ligament and anterior cruciate ligament scores in donors younger than 45 years (n=9) with normal cartilage (grade 0). This figure is only reproduced in colour in the online version.

In order to differentiate between changes due to aging and those due to osteoarthritis-like cartilage changes we divided knees from donors older than 60 years into two groups. The first group—cartilage grades I–II (n=66) represented the changes occurring during normal aging; while the second group with cartilage grades III–IV (n=29) represented knees with osteoarthritis. Of the 66 knees with cartilage grades I–II the PCL histological grade was classified as 0% normal, 74% mild, 15% moderate and 11% severe. The ACL from the same knees were classified as 2% normal, 45% mild, 14% moderate and 12% severe. This indicates that during the aging process both ligaments have equal severity scores. On the other hand, of the 29 knees with cartilage grades III–IV the PCL histological grade was classified as 69% mild, 14% moderate and 17% severe. ACL from the same knees were classified as 45% mild, 28% moderate and 28% severe grade. Therefore, during arthritis progression more histopathological changes are observed in the ACL than in the PCL.


Although disruption and loss of the articular cartilage is a hallmark of knee osteoarthritis, the disease process results in changes in other knee tissues such as ligaments, menisci, subchondral bone and synovial membrane.24–27 Moreover, there is the unresolved question of the differences in patterns of tissue changes taking place during normal aging versus patterns observed in osteoarthritis. To understand better the changes in the cruciate ligaments with aging and osteoarthritis, we conducted a comprehensive analysis of major histological patterns of PCL pathology in 120 knees across the adult age spectrum. Overall, our findings indicate that PCL pathology in the context of knee osteoarthritis is very distinct from ACL pathology.

The study has several limitations: it is a cross-sectional study in which there is a risk of reverse causality. Although there was statistical significance between the different osteoarthritis groups, the sample size between the groups was not equal and was determined by donor availability. Due to multiple testing there is a potential for low power. Some donor information, such as the extent of lower extremity malalignment, was not available.

PCL rupture

In the present postmortem study all PCL were intact and the ACL were ruptured in 10% of the knees. Loss of the ACL in severe osteoarthritis is attributed to mechanical impingement28 and ligament abrasion against osteophytes.18 As the PCL is not typically subjected to impingement and abrasion in knee osteoarthritis, this may partly account for the less severe changes.

PCL degeneration and aging

The PCL begins to degenerate early but does not continue to degenerate with aging. PCL changes, particularly fibre disorganisation, mucoid degeneration and chondroid metaplasia, were the earliest to occur, even before degeneration in the articular cartilage was observed. The PCL histopathological scores were higher than the ACL scores in donors up to the age of 50 years, beyond which ACL scores became more severe. This suggests that PCL changes reflect a milder degenerative process, which is not profoundly accelerated by aging. The severity of the PCL histological changes also did not correlate with the age of patients with end-stage knee osteoarthritis.29 Over the age of 50 years, PCL degeneration is more associated with the presence of cartilage damage rather than age.

Two groups of knees demonstrated distinct patterns of PCL pathology. Knees from patients older than 60 years with high-grade cartilage damage had significantly worse total PCL scores compared to knees in the same age group with lower-grade cartilage damage. The second subset was knees with ruptured ACL. In those knees, the PCL scores were significantly higher than in knees with intact ACL. The divergent histopathological changes in the ACL and the PCL indicate independent pathological mechanisms.

PCL histology and correlation to cartilage degeneration

The PCL histological scores were lower in knees with cartilage grade IV compared to grade III, resulting in a poor correlation with cartilage grade. In contrast, the ACL histological scores from the same donors increased continuously with increasing cartilage grades. Moreover, there was no correlation between the total PCL score and the cartilage scores of the specific knee compartments. ACL tears have been associated with an increase in lateral compartment osteoarthritis,14 presumably due to shift in the contact stress posteriorly and laterally on the surface of the tibia on MRI and fluoroscopy.30 In dogs, transection of the PCL resulted in less severe articular cartilage degeneration and osteophyte formation than transection of the ACL.31 PCL degeneration may contribute less to overall joint pathology than ACL degeneration.

Histological changes in PCL versus ACL

Changes in collagen fibrils affect the biomechanical properties of the ligament,32–34 and there is a decrease in collagen fibrils in the PCL in individuals older than 60 years.35 In our study, fibre disorganisation was the earliest and most common change in the extracellular matrix. It was also the most common histopathological change in patients younger than 45 years with normal-appearing cartilage and was more common in the PCL than the ACL. It was also the only histopathological change in the PCL that increased continuously with the changes in the articular cartilage. In contrast, there was a weak correlation between PCL fibre disorganisation and donor age. The ACL exhibited overall higher levels of fibre disorientation, which correlated with donor age.

Chondroid metaplasia, a shift in ligament cell phenotype towards a more chondrocytic morphology and cartilaginous matrix36 ,37 was noted in 75% of the PCL. Mucoid changes represent degeneration of collagen and replacement with glycosaminoglycans.38 While clinically symptomatic, mucoid degeneration is a rare condition.39 It has previously been described histologically in patients undergoing knee arthroplasty,13 ,40 and been visualised by MRI in a small number of cases.41 Based on the present findings, mucoid degeneration appears to develop gradually, first appearing in the fifth decade, with minimal increase in severity thereafter. Overall, PCL inflammation scores were lower than all other histopathological changes and correlated little with age, osteoarthritis grade, or any other histopathological measure. Inflammation was lower in the PCL relative to the ACL in all the age groups and in all levels of cartilage degeneration. This suggests that the inflammation in the PCL is not likely to be a significant factor in aging, degeneration, or osteoarthritis.


Lilo Creighton and Melissa Szeto are gratefully acknowledged for their support for the histology of the specimens.


View Abstract
  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors All authors were involved in drafting the manuscript and all authors approved the final version. MKL had full access to all data and takes responsibility for the integrity of the data and the accuracy of the data analysis. DDD, MKL, YDL, JAK: study conception and design; YDL, AH, SP: acquisition of data; JAK, YDL, MKL, DDD: analysis and interpretation of data.

  • Funding This study was supported by the National Institutes of Health (AG007996), Donald and Darlene Shiley, and by an Osteoarthritis Research Society International Scholarship Award (YDL).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.