Regional variations in human patellar trabecular architecture and the structure of the proximal patellar tendon enthesis

Abstract

Proximal patellar tendinopathy occurs as an overuse injury in sport and is also characteristic of ankylosing spondylitis patients. It particularly affects the posteromedial part of the patellar tendon enthesis, although the reason for this is unclear. We investigated whether there are regional differences in the trabecular architecture of the patella or in the histology of the patellar tendon enthesis that could suggest unequal force transmission from bone to tendon. Trabecular architecture was analysed from X-rays taken with a Faxitron radiography system of the patellae of dissecting room cadavers and in magnetic resonance images of the knees of living volunteers. Structural and fractal analyses were performed on the Faxitron digital images using MatLab software. Regional differences at the enthesis in the thickness of the uncalcified fibrocartilage and the subchondral plate were evaluated histologically in cadaveric material. The radiological studies showed that the quantity of bone and the apparent trabecular thickness in the patella were greatest medially, and that in the lateral part of the patella there were fewer trabeculae which were orientated either antero-posteriorly or superiorly inferiorly. The histological study showed that the uncalcified fibrocartilage was most prominent medially and that the subchondral plate was thinner laterally. Overall, the results indicate that mechanical stress at the proximal patellar tendon enthesis is asymmetrically distributed and greater on the medial than on the lateral side. Thus, we suggest that the functional anatomy of the knee is closely related to regional variations in force transmission, which in turn relates to the posteromedial site of pathology in proximal patellar tendinopathy.

Introduction

Overuse injuries affecting the patellar tendon entheses are well documented in a variety of sports. Both ends of the tendon are vulnerable, but the most common site of pathology is near the posteromedial portion of the patella (Yu et al. 1995; Khan et al. 1999) – a condition sometimes known as ‘jumper's knee’. However, in skeletally immature patients, the tibial insertion of the patellar tendon is the site of an inflammatory disorder known as Osgood Schlatter's syndrome and both entheses are affected in patients who suffer from ankylosing spondylitis (Balint et al. 2002).

Epidemiological studies suggest that proximal patellar tendinopathy occurs in many athletic activites and is not restricted to jumpers, as the term ‘jumper's knee implies (Warden & Brukner, 2003). Nevertheless, it is common in elite volleyball and basketball players (Ferretti et al. 1990; Cook et al. 1998, 2000; Lian et al. 2005) and athletes who are capable of high vertical jumps are at particular risk of developing patellar tendinopathy (Lian et al. 1996). It is unclear why the pathological process mainly affects the posteromedial part of the proximal tendon enthesis. It is not known whether the problem primarily relates to a unique anatomical feature of the enthesis in this region, to an unequal pattern of force transmission between the patella and the patellar tendon, or both. In order to evaluate these possibilities, we have described regional differences in the trabecular architecture of the patella and in the histological structure of the proximal patellar tendon enthesis.

Materials and methods

Faxitron radiology

Fifteen dissecting room cadavers (ten male, five female, ages 74–101 years) donated to the Cardiff University for anatomical examination under the provision of the 1984 Anatomy Act and the 1961 Human Tissue Act were used. The cadavers had been perfused for 72 h with an embalming fluid containing 4% formaldehyde and 25% alcohol. They were selected according to the quality of preservation and the absence of gross abnormalities in the knees. Medical histories (other than the cause of death) were not available. Both patellae (right and left) were removed from each cadaver by cutting the quadriceps and patellar tendons transversely, approximately 1 cm from the bone. The left patella was divided longitudinally into medial, central and lateral parts and the right patella was divided transversally into distal, intermediate and proximal regions. This enabled us to analyse bone structure in nine areas (M1, M2, M3, C1, C2, C3, L1, L2 and L3) as shown in Fig. 1(a). Areas M1–M3 together formed a single medial facet, but each of the other areas represented equivalent separate facets. The use of both longitudinally and transversely sectioned patellae enabled us to examine trabeculae in all three standard planes (anterior–posterior, lateral–medial and superior–inferior). Standard X-rays were taken on a Faxitron Specimen Radiography System (Model MX-20, Wheeling, IL, USA) with a high-resolution algorithm. The imaging parameters were identical for all the specimens and the following settings were used for taking the contact radiographs: 300 s, 0.3 mA and 35 kV; pixel matrix, 1024 × 1054; resolution 200 pixels mm⁻¹. Images were transferred to a personal computer and structural analyses of the bone were performed. Algorithms used to characterize bone architecture were all developed in our laboratory using Matlab Software (Version 6.10.450, release 12.1).

Figure 1. Gross anatomy of the right patella: posterior views. (a) The different areas of the bone in which trabecular architecture was analysed. Nine areas were studied (M1–3, C1–3 and L1–3), defined by two vertical and two horizontal lines. The right vertical line (arrow) lay along the vertical ridge that separated an area of the patella which articulated with the lateral femoral condyle from one which articulated with the medial condyle. (b) The attachments of the quadriceps and patellar (PT) tendons to the patella and the common site of proximal patellar tendinopathy (*). At the superior border of the patella, the quadriceps tendon is dominated by the attachments of vastus intermedius and rectus femoris (VI-RF). Note that the oblique fibres of vastus medialis (VM) have an extensive attachment to the medial border of the patella, whereas the lateral border is largely devoid of muscle attachment (arrow). Note also that there is a direct connection between the vastus medialis tendon and the patellar tendon (arrowheads). (c) An enlarged view of the connection between the vastus medialis tendon and the patellar tendon (arrowheads). P, patella.

Matlab treatment

The procedures used have been described previously by Cortet et al. (2002, 2004). The boundary between the thin cortical bone and the underlying trabecular bone was defined using an automatic contour detection algorithm. Subsequently, a segmentation process permitted separation of spicules from bone marrow (Fig. 2). The segmentation was made with an edge detection using a LaPlacian–Gaussian filter. This includes both a smoothing filter (which convolutes the image by a Gaussian filter) and a second-order derivative filter. Tuning of this combined filter addresses the size of the smoothing window, but also the variance of the convolutive Gaussian filter. Zero-crossing detection in the resulting image provides a binary image in which dark regions represent the bone marrow and light regions represent trabeculae. The different steps of image analysis are illustrated in Fig. 2(c–g). Several variables were calculated on the binary images based on two-dimensional (2D) analysis (Cortet et al. 2004). These were (1) apparent bone volume (BV) = Trabecular bone (TB)/tissue volume (TV), (2) apparent trabecular number (TN) = (BV/TV)/TB, (3) trabecular separation (TS) = (1/TN) − TB and (4) apparent trabecular thickness (TH) = 2/(TS/BV). It should be noted that these variables were obtained by analysing 2D X-ray images. As this cannot equate to the real properties of 3D bone, it is the comparative rather than the absolute values of each parameter in different regions of the patella which are important in our study. In recognition of this, we have followed the convention of previous workers in referring to the various bone parameters as ‘apparent’ rather than ‘real’ (Cortet et al. 2002, 2004). In addition, we validated our analysis on the X-ray images by conducting a preliminary study in which we compared trabecular thickness, number and separation in histological sections of the patella with those obtained from X-ray images. The coefficient of correlation was always greater than 0.95.

Figure 2. Faxitron analyses of the patella. (a) A para-sagittal section through the medial part of the right patella, showing prominent trabeculae (arrows) running from superior (S) to inferior (I). A, anterior; P, posterior. (b) A transverse section of a left patella showing prominent trabeculae running from medial (M) to lateral (L) (arrows), and from anterior to posterior (arrowheads). (c–g) The different stages in the treatment of images obtained from a MatLab analysis. (c) An example of a binary transformation of trabeculae. (d) A skeletonized image of the binary transformation shown in (c). (e–g) Trabecular architecture derived from skeletonised images showing trabeculae running in horizontal (e), vertical (f) and oblique (g) directions.

Trabecular orientation was studied using a discrete, 2D transformation and a stationary wavelet transformation (SWT). The SWT performs a multi-level 2D stationary wavelet decomposition using specific orthogonal wavelet filters. This type of SWT (Nason & Silverman, 1995) leads to a decomposition of approximation in three orientations [horizontal (Fig. 2e), vertical (Fig. 2f) and oblique (Fig. 2g)].

Magnetic resonance imaging (MRI)

Three healthy male volunteers (31–60 years) were studied on a 1.5-T magnetic resonance scanner (Siemens, Erlangen, Germany). Multi-slice, T1-weighted non-fat-saturated conventional spin-echo images (TR = 500 ms, TE = 16 ms, 1 mm slice thickness, 448 × 384 matrix, field of view 107 mm) were taken in the sagittal, coronal and transverse planes using a flexible surface coil placed anterior to the patella.

Histology

For examining the proximal patellar tendon enthesis, tissue from ten female cadavers was post-fixed in 10% neutral buffered formalin, decalcified with 5% nitric acid, dehydrated through a graded alcohol series, cleared in xylene and embedded in paraffin wax. Serial longitudinal sections were cut at 8 µm throughout the medial, central and lateral thirds of the enthesis and 12 sections were mounted on glass slides at 1-mm intervals. Slides were stained with Hall and Brunt's quadruple stain (Hall, 1986), Masson's trichrome (for photography) and toluidine blue (for fibrocartilage metachromasia).

Morphometric analysis of histological sections

At all three parts (medial, central and lateral) of the enthesis, the thickness of the subchondral plate was assessed at each 1-mm sampling point by measuring the distance from the tidemark to the distal part of the plate at 15 sites, equally spaced at approximately 500 µm along the section. The thickness of the zone of uncalcified fibrocartilage was estimated by measuring the distance from the tidemark to the furthest recognizable chondrocyte within the tendon, following a protocol adopted previously (Benjamin et al. 1991). Measurements were made with a micrometer eyepiece at a magnification of ×100. Five such measurements were taken at equal intervals across the attachment site on one slide at each 1-mm sample point.

Statistical analysis

Mean values reported are given ± standard deviations (SD). A repeated-measures analysis of variance (anova) with Fisher's post-hoc tests was used to compare medial, central and lateral parts of the patella for each variable. The P-value for significance was set at 0.05.

Results

The gross anatomy of the attachment of the quadriceps and patellar tendons to the patella is shown in Fig. 1. Note that the oblique fibres of vastus medialis are not only attached to the medial border of the patella but also have a small region of direct continuity with the patellar tendon (Fig. 1b,c).

Faxitron radiological analysis

A simple visual inspection of the radiographs showed that there were clear differences in the density and arrangement of trabecular bone in the patella, from proximal to distal and from medial to lateral. This was confirmed by algorithm analyses of the X-ray images. The quantitative data were used to facilitate the objective description of regional differences in bone structure within the patella. The results are summarized below in Figs 5–7 and the most significant findings are detailed below. Essentially, in the proximal and distal regions of the bone, the apparent trabecular volume fraction was significantly higher centrally and medially than it was laterally (Fig. 5). However, in the intermediate region, there was no significant difference from medial to lateral (Fig. 5). There were similar differences in apparent trabecular thickness (Fig. 6). Thus, trabeculae were thickest in the medial and central parts of the proximal and distal regions, but there was no difference in thickness in the intermediate region from the medial to the lateral aspects of the patella. The apparent number of trabeculae in the intermediate region was significantly less than in the proximal and distal regions (Fig. 7). Significant differences between the values for the facets in each part of the patella were only found laterally – facet L1 differed significantly from both L2 and L3.

Figure 5. Apparent trabecular bone volume fraction in the patellar facets expressed as a percentage of the total tissue volume (mean values ± SD). Significant differences are underlined. The black columns show, respectively, from left to right, the average (A) of the individual values for the lateral (L1, L2, L3), central (C1, C2, C3) and medial (M1, M2, M3) sides.

Figure 6. Apparent trabecular thickness in the different facets (mean values ± SD). Significant differences are underlined. The black columns show, respectively, from left to right, the average (A) of the individual values for the lateral (L1, L2, L3), central (C1, C2, C3) and medial (M1, M2, M3) sides.

Figure 7. Apparent trabecular number in the different facets (mean values ± SD). Significant differences are underlined. The black columns show, respectively, from left to right, the average (A) of the individual values for the lateral (L1, L2, L3), central (C1, C2, C3) and medial (M1, M2, M3) sides.

Trabeculae that ran from the anterior to the posterior surface of the patella and from its superior to its inferior pole were strikingly evident in the sagittal plane in both X-ray and MRI images (Figs 2a and 3e). However, it is also clear that trabeculae ran in other directions, which are collectively referred to as ‘oblique’ planes. The anterior–posterior and oblique trabeculae were also a feature of transversely sectioned patellae, but this plane of section also highlighted spicules that were orientated from medial to lateral (Fig. 2b). By combining information from patellae sectioned in both planes, we were able to analyse the trabecular architecture in all three dimensions. In Tables 1 and 2, the number of trabeculae in any one plane is expressed as a percentage of the total trabeculae detected in that patella. The analysis of longitudinally sectioned patellae shows that there were a significantly greater number of trabeculae orientated antero/posteriorly in the medial and central parts of the bone than laterally (Table 1), but the number of trabeculae orientated obliquely or in the medial–lateral planes was similar in all three regions. The analysis of transversely sectioned patellae shows that there were significantly more trabeculae orientated either superiorly inferiorly or anteriorly posteriorly in the medial and central parts than in the lateral part (Table 2). By contrast, the number of trabeculae aligned obliquely was significantly higher laterally.

Figure 3. MRI scans of the patellar tendon. (a) A sagittal image of a knee in full extension. The scan passes through the region of the vertical ridge separating the medial and lateral facets. Note the continuity of the patellar (PT) and quadriceps (QT) tendons anterior to the patella (P) and the position of Hoffa's fat pad (HP) adherent to the posterior surface of the patellar tendon. F, femur. (b) An image taken in the transverse plane near the superior part of the patellar tendon close to its proximal enthesis. Note how Hoffa's pad infiltrates and interdigitates with the posterior surface of the tendon (arrows). (c,d) Sagittal images from the medial (c) and lateral (d) sides of the patella showing the regional difference in the extent to which Hoffa's pad infiltrates (arrow) the patellar tendon. The infiltration is greater on the medial than on the lateral side. (e,f) Sagittal images from the medial (e) and lateral (f) sides of the patella showing the difference in the prominence of the superior–inferior trabeculae (arrows). The trabeculae are more prominent medially than they are laterally. (g) A frontal view of a patella showing the prominence of the superior–inferior trabeculae within the central part of the bone (arrows).

Table 1.  Mean (± SD) apparent trabecular number for the anterior/posterior, lateral/medial and oblique trabeculae obtained from the transversely sectioned right patellae

	Facets	L/M	A/P	Oblique
* Significantly different from the lateral value (P < 0.05).
Proximal region	L1	34.0 ± 3.9	22.3 ± 4.2	43.8 ± 5.8
	C1	35.9 ± 6.4	33.2 ± 5.5	31.0 ± 6.2
	M1	38.0 ± 4.4	29.0 ± 4.3	33.0 ± 5.2
Intermediate region	L2	39.8 ± 6.6	25.2 ± 4.7	35.0 ± 5.3
	C2	35.0 ± 7.1	32.1 ± 5.6	32.0 ± 6.7
	M2	34.0 ± 5.2	34.0 ± 6.3	32.0 ± 7.1
Distal region	L3	42.0 ± 4.5	24.4 ± 3.9	33.7 ± 4.9
	C3	37.6 ± 6.6	25.8 ± 5.2	36.5 ± 4.7
	M3	36.0 ± 4.1	30.0 ± 5.2	34.0 ± 5.4
Average lateral (L1, L2 and L3)		38.6 ± 5	24.0 ± 4.3	37.5 ± 5.3
Average central (C1, C2 and C3)		36.2 ± 6.7	30.4 ± 5.4*	33.2 ± 5.9
Average medial (M1, M2 and M3)		36.0 ± 4.6	31.0 ± 5.3*	33.0 ± 5.9

Table 2.  Mean (± SD) apparent trabecular number for the anterior/posterior, vertical and oblique trabeculae obtained from the longitudinal section for the patellar facets

	Facets	Vertical	A/P	Oblique
* Significantly different from the lateral values P < 0.05. ** Significantly different from the central and medial values, P < 0.05.
Proximal region	L1	11.2 ± 5.2	39.1 ± 5.1	49.7 ± 6.3
	C1	27.6 ± 5.1	51.3 ± 6.2	21.1 ± 4.5
	M1	29.5 ± 4.9	45.2 ± 5.4	25.3 ± 5.7
Intermediate region	L2	22.9 ± 5.6	40.2 ± 5.2	36.9 ± 6.1
	C2	26.1 ± 4.2	52.1 ± 4.6	21.8 ± 7.1
	M2	26.2 ± 4.4	43.8 ± 3.9	30.9 ± 5.2
Distal region	L3	21.7 ± 5.5	29.3 ± 6.8	49.9 ± 4.4
	C3	29.6 ± 5.3	49.1 ± 6.6	21.5 ± 5.6
	M3	25.1 ± 4.2	49.2 ± 5.5	25.7 ± 4.2
Average lateral (L1, L2 and L3)		18.6 ± 5.4	36.2 ± 5.7	45.5 ± 5.6**
Average central (C1, C2 and C3)		27.8 ± 4.9*	50.8 ± 5.8*	21.5 ± 5.7
Average medial (M1, M2 and M3)		26.9 ± 4.5*	46.1 ± 4.9*	27.3 ± 5.0

Magnetic resonance imaging analysis

The MRI observations show that the quadriceps and patellar tendons are continuous with each other over the anterior surface of the patella (Fig. 3a). This feature is more obvious medially than laterally. The deep surface of the patellar tendon is closely related to Hoffa's fat pad across its entire width (Figs 3a,b and 4) and the fat pad forms distinctive interdigitations with the proximal part of the tendon that are clearly visible in T1-weighted images (Fig. 3b). The interdigitations were more obvious on the medial than on the lateral side (Fig. 3c,d).

Figure 4. A low-magnification view of a histological section of the inferior pole of the patella (P) showing the enthesis (E) of the patellar tendon (PT). Note the conspicuous fibres (*) that pass anterior to the patella rather than attaching to its inferior pole, and the trabeculae that run both in a superior–inferior (T1) and an anterior–posterior (T2) direction. Hoffa's fat pad (HP) is intimately related to the deep surface of the patellar tendon and has fibrous strands coursing through it which blend with the tendon itself (arrow). Scale bar = 3 mm. Masson's trichrome.

As with results from the Faxitron analysis, both the quantity of bone and the orientation of the trabeculae differed according to site. Although the MRI observations were purely qualitative, they confirmed that there were more trabeculae running antero-posteriorly or superiorly-inferiorly in the medial than in the lateral part of the patella (Fig. 3e,f). It was also clear that there were more trabeculae running superiorly-inferiorly in the central part of the patella (Fig. 3g).

Histology

The histological sections confirmed the MRI observation that Hoffa's fat pad was adherent to the tendon. The fat contained bundles of collagen fibres that blended with the deep surface of the tendon (Fig. 4).

The morphometric analysis showed that the thickness of the subchondral plate was significantly greater (P < 0.05) in the central part of the enthesis than medially (Figs 8a,b and 9). However, there was no significant difference either between the central and lateral parts or between the lateral and medial parts (Fig. 9). The zone of uncalcified fibrocartilage was significantly thicker medially than laterally (Figs 8c,d and 9).

Figure 8. Histological sections of the proximal patellar tendon enthesis. (a,b) Differences in the thickness of the subchondral bone plate (SP) in the medial (a) and central (b) parts of the enthesis. Note that the subchondral plate is thicker centrally. TM, tidemark. (c,d) Differences in the quantity of uncalcified fibrocartilage (UF) in the medial (c) and lateral (d) parts of the enthesis. Note that the fibrocartilage is more conspicuous medially. DT, dense fibrous connective tissue. Scale bars = 200 µm. Masson's trichrome.

Figure 9. The thickness of the subchondral bone plate and the zone of uncalcified fibrocartilage in the lateral (L), central (C) and medial (M) parts of the proximal patellar tendon enthesis (mean values ± SD). Significant differences (P < 0.05) are underlined.

Discussion

It is generally accepted that changes in force transmission through cancellous bone result in a modification of its architecture. This is the principle underpinning Wolff's law. Although a quantitative relationship between strain levels and changes in bone parameters has yet to be documented, it is widely assumed that analysis of trabecular bone architecture provides valuable information on stress patterns within cancellous bone (e.g. Pal & Routal, 1998). However, it could be argued that when bone is subject to altered load, it adapts by changing its trabecular architecture so that there is no overall change in stress levels. Yet this does not explain the common observation that osteoporotic or osteopenic bone has a reduced resilience to traumatic insult that is associated with an increased incidence of osteoporotic vertebral compression fractures in the elderly (Patel et al. 1991). Furthermore, stress fractures that are linked to altered stress patterns can occur in non-osteoporotic bones in young adults active in sport. These include navicular fractures and stress fractures of the patella (Orava et al. 1996; Saxena et al. 2001).

Note that the material used for the Faxitron radiographic analysis of trabecular architecture came from dissecting room cadavers whose activity levels in life were unknown. All were elderly individuals and likely to be less active (and perhaps have a different gait) than a younger cohort at risk of developing patellar tendinopathy. Some were osteoporotic. It was for these reasons that the cadaveric studies were complemented with an MRI analysis of the knees of younger individuals. It should also be noted that for ethical and/or practical reasons, it was not possible to obtain tissue samples from patients suffering from proximal patellar tendinopathy. All these points should be borne in mind when evaluating the significance of our findings in relation to proximal patellar tendinopathy. The validity of the quantitative approach to comparing the characteristics of trabecular bone in different regions of the patella was confirmed by rigorous preliminary experiments and for comparative purposes the data were normalized per unit volume. The detailed analysis that we have presented of regional variations in the trabecular architecture of the patella from proximal to distal and from medial to lateral allows us to analyse the pattern of force transmission from quadriceps to patellar tendons. We suggest that the lines of force transmission have an important bearing on the regional vulnerability of the posteromedial part of the proximal patellar tendon enthesis to enthesopathy.

In the proximal region of the patella, the total amount of bone tissue and the thickness of the trabeculae were significantly greater in the medial and central parts than in the lateral part. Furthermore, trabecular orientation differed according to site. Laterally, there were fewer trabeculae running either antero-posteriorly or superiorly inferiorly than there were medially or centrally. Collectively, the data suggest that force transmission and mechanical stress in the proximal region of the patella are asymmetrically distributed – forces are greater medially and centrally than laterally. Thus, the sites of greatest force transmission are at the insertions of rectus femoris and vastus medialis, and the relative paucity of bone laterally at the proximal pole of the patella reflects the lack of any direct muscle insertion in this region. Vastus lateralis is mainly inserted into the lateral retinaculum rather than directly into the patella itself, while the oblique fibres of vastus medialis attach to the medial border of the patella (Standring, 2004). The inequality of force transmission could relate to an imbalance in the mechanism of attachment of the quadriceps tendon to the patella, muscle activation or muscle stiffness. This is in line with the suggestion of Mariani et al. (1978) that jumper's knee is associated with an imbalance of the knee extensor muscles.

Although vastus medialis inserts obliquely onto the medial border of the patella, our results show that the trabeculae in this region are predominantly orientated sagitally, as at the attachment of rectus femoris. Such an arrangement of trabeculae can be explained by the changing lines of force transfer between the quadriceps tendons, patella and patellar tendon. In full extension, mechanical stress is largely transmitted supero-inferiorly from the quadriceps to the patellar tendon, via the patella. However, as the knee moves towards full flexion, some of the force is transferred from the patella to the patellar groove of the femur. This explains the presence of trabeculae orientated in an antero-postero direction.

In the intermediate portion of the patella (i.e. half way between the proximal and distal poles), the variability in trabecular characteristics from medial to lateral is no longer evident and the total quantity of bone tissue is greater than it is proximally in any of the regions defined (i.e. medial, central or lateral). Collectively, the results for the intermediate portion of the patella suggest that the resultant pull of vastus medialis and rectus femoris has now been dissipated across the bone, allowing the patella to resist compression against the femur and contributing to the positional stability of the bone. The compression force and the mechanical stress acting on the patella vary with the angle of knee flexion and are greatest at higher flexion angles. Repetitive cycles of high compressive force resulting from flexion–extension movements probably account for the greater quantity of bone tissue that was observed in the intermediate region and the high proportion of trabeculae orientated antero-posteriorly. The homogeneity observed in trabecular characteristics in the intermediate region of the patella suggests that the compressive force was equally distributed from medial to lateral. The uniform compressive force contributes to the stability of the patella and contributes to maintaining the alignment of the quadriceps tendon, patella and patellar tendon when the quadriceps muscles are activated and/or the knee is flexed. This is further enhanced by the considerable number of fibres derived from both the quadriceps and the patellar tendons that pass anterior to the bone. They link the two tendons together and facilitate the attachment of both to the anterior surface of the patella, bracing it against the femur in the process. An extensive attachment of the patellar tendon to the anterior surface of the patella has also been noted by Basso et al. (2001).

In the distal region of the patella, the trabecular variability that was evident proximally, but disappeared in the intermediate region, is now re-established. Thus, the medial and central parts have a trabecular architecture that is similar to each other, but in both cases the total quantity of bone tissue is greater than it is on the lateral side. We interpret this to mean that less force is transmitted from the patella to the patellar tendon laterally. This helps to explain why the lateral side of the tendon appears to be relatively protected from patellar tendinopathy. There is clearly a stress difference at the distal pole of the patella which reflects the areas most at risk of patellar tendinopathy, i.e. the medio-central part of the patellar tendon origin. Thus, the patellar tendon is not acting as a uniform transmitter of force in the sagittal plane, but has a differential stress transmission that concentrates more on the central and medial areas than on the lateral. This in turn relates to the requirement of the patella − patellar tendon complex to resist lateral forces which are attempting to subluxate the patella laterally because of the existence of the Q angle (the quadriceps angle, i.e. the angle between a line which is drawn linking the anterior superior iliac spine to the centre of the patella and a second line drawn from the latter point to the middle of the tibial tuberosity). This is in keeping with the greater thickness of the medial para-patellar retinacular fibres compared with the rather flimsy nature of the retinacular fibres on the lateral side (J. A. Fairclough, unpublished observations). At the distal pole of the patella, the greater number of trabeculae orientated either superiorly inferiorly or anteriorly posteriorly in the medial and central parts compared with the lateral part suggests that the composite force is greatest in the proximal, centro-medial portion of the patellar tendon. This corresponds exactly to the area most vulnerable to patellar tendinopathy (Hamilton & Purdam, 2004). It also suggests that the occurrence of patellar tendinopathy at the site of maximum stress is affected by the relative pull of the vastus medialis. This implies that muscle dysfunction may be the primary causative agent in the aetiology of patella tendonitis (Witvrouw et al. 2001). However, it must also be recognized that the intimate relationship between the patellar tendon and Hoffa's pad, which was particularly obvious posteromedially in our MRI scans, cautions against the view that the symptoms of patellar tendinopathy are purely related to stress concentration within the tendon. Indeed, Sanchis-Alfonso et al. (2001) have raised the possibility that neural changes in Hoffa's pad could contribute to the pain associated with jumper's knee. Hoffa's pad does form a functional unit with the patellar tendon and that the fibrous strands running through it could be regarded as a part of the tendon itself. Thus, the possibility needs also to be considered that Hoffa's pad has a mechanical role in stress dissipation at the proximal patellar tendon enthesis and that it forms part of an ‘enthesis organ complex’ with the osteotendinous junction itself (Benjamin & McGonagle, 2001; Benjamin et al. 2004).

The histological analysis of the proximal patellar tendon enthesis complements the radiological data. The finding of a greater quantity of uncalcified fibrocartilage medially than laterally suggests a higher level of stress concentration medially. It is thus intriguing that the quantity of fibrocartilage is greater in patients with proximal patellar tendinopathy (Ferretti et al. 1983) and that there are also reports of an increase in mucoid degeneration in these patients (Khan et al. 1996; Cook et al. 1997; Panni et al. 2000). This presumably reflects the greater stress concentration on the enthesis in symptomatic knees. Fibrocartilage can dissipate stress concentration because it has different physical properties from the dense fibrous connective tissue that is typical of tendons (Woo et al. 1988; Benjamin & Ralphs, 1998). Its transitional character helps to balance the different elastic moduli of tendons and bone (Hems & Tillmann, 2000). It should be regarded as a tissue that is primarily adapted to resisting compressive and/or shear forces rather than tensile load (Benjamin & Ralphs, 1998). Thus, its prominence in the posteromedial part of the patellar tendon supports the suggestion of Almekinders et al. (2001) and Hamilton & Purdam (2004) that proximal patellar tendinopathy may result from compression of the patellar tendon against the femur. Almekinders et al. (2001) argue that a tensile load in the anterior fibres of the patellar tendon produces a stress-shielding response in the posterior fibres. These are compressed against the femur when the knee is flexed.

In conclusion, our study demonstrates that there is a variation in the trabecular architecture of the patella and in the structure of the proximal enthesis of the patellar tendon. This can explain the common posteromedial location of proximal patellar tendinopathy. We have shown that the trabecular architecture in the patella varies in a complex manner from one facet to another. As Wolff's law dictates that trabecular characteristics reflect the lines of force transmission and mechanical stress within bones, we suggest that mechanical stress at the proximal patellar tendon enthesis is asymmetrically distributed. Stress is greater medially than laterally and this is in line with the particular site of pathology in proximal patellar tendinopathy.

Acknowledgement

This work was supported by Action Medical Research and Search.