Journal of Orthopaedic Research 9559-567 Raven Press, Ltd., New York 0 1991 Orthopaedic Research Society
Effects of Osteochondral Defect Size on Cartilage Contact Stress Thomas D. Brown, David F. Pope, Joseph E. Hale, Joseph A. Buckwalter, and Richard A. Brand Biomechanics Laboratory, Department of Orthopaedic Surgery, The University of Iowa, Iowa City, / A , U.S.A.
Summary: Contact stress distributions were studied in vitro for 13 dog knees, with full-thickness osteochondral defects drilled in the weight-bearing area of both femoral condyles. Diameters of the circular defects were concentrically enlarged from 1 to 7 mm. Digitally-imaged Fuji film was used to record cartilage contact stress distribution on femoral condyles for each increment of defect diameter. All specimens showed at least some tendency for contact stress concentration at the rim of the defects. However, detailed distributions had large interspecimen variability and, within a given specimen, contact stress distributions became progressively more nonuniform around the defect rim as the diameter was enlarged. Averaged over the full series of 26 condyles, circumferential mean cartilage contact stress around the defect rim was only moderately higher (by 10-30%) than intact surface’s peak local contact stress [series average = 6.2 mega pascals (MPa)]. Maximal rim stress concentration occurred for 2 mm defects, there being a consistent trend toward mild rim stress decrease with further defect enlargement. Such modest contact stress elevations, per se, are probably insufficient to inhibit defect repair or to cause degeneration of surrounding cartilage. However, near the defect rim (for all diameters), the radial component of the gradient of contact stress (i.e., radialdirection variation of contact stress) was consistently elevated by an order of magnitude above that for intact, condyle articular cartilage. Key Words: Contact stress-Cartilage healing-Cartilage repair-Osteochondral defect.
While superficial articular cartilage defects possess little intrinsic capacity for repair (7,8,17,18), lesions extending through the subchondral bone show variable ability to repair (3,5,6,9-11,19,20, 22,24,26). Smaller lesions (i.e., up to 1-3 mm, depending upon species) tend to heal, sometimes with almost normal-appearing hyaline-like cartilage (24), while larger lesions (2-3 mm or more) tend to heal with fibrocartilage of variable quality (6,9,22). Clinically, the presence of a significant chondral defect
(i.e., clinically detected osteochondritis dissecans) is associated with a much higher incidence of osteoarthrosis than occurs in the general population, and clinically significant disease appears approximately a decade earlier than with idiopathic arthrosis (16). This clearly implies that some phenomenon, associated with defects, leads to arthrosis. Larger osteochondritic defects are associated with poorer clinical outcomes than are smaller ones (15), suggesting, perhaps, that some size-related mechanical phenomenon explains degeneration, as it might the quality of early repair. The reason (or cause) for variable repair is unclear. Based on substantial indirect evidence, Radin and Burr (23) suggested that while a mechanical
Received May 16, 1990; accepted September 21, 1990. Address correspondence and reprint requests to Dr. T. D. Brown, Department of Orthopaedic Surgery, The University of Iowa, Iowa City, IA 52242, U.S.A.
559
T. D . BROWN ET AL.
560
stimulus is essential for normal cartilage maintenance and repair, the magnitude of that stimulus is not large. They also suggested that reduction of local stress concentrations around defects favors repair. One could extend such reasoning and postulate that larger lesions heal poorly because their surrounding cartilage is subjected to deleteriously high contact stresses (i.e., contact force per unit surface area). In a recent in vivo canine study (22), we found that the highly compliant, fibrocartilaginous repair tissue, filling a large (6 mm) femoral condyle osteochondral defect, is incapable of moderating appreciable contact stress disruptions acutely present after defect creation. A related issue of clinical importance is the relationship between defect size and putative mechanically induced degeneration of surrounding cartilage. At present, however, the degree to which defect size, per se, causes local contact stress concentrations is not known. A study was, therefore, designed to test the hypothesis that cartilage surrounding large-diameter osteochondral defects experiences substantially elevated contact stress compared to that experienced around otherwise equivalent, small-diameter defects. MATERIALS AND METHODS Articular contact stress distributions were mapped around full-thickness osteochondral defects, created on both femoral condyles of 13 adult, mongrel dog knees. All specimens were excised, at necropsy, from 20-25 kg animals that had been normally active, prior to sacrifice, in the course of an unrelated acute study. Following hip disarticulation, all soft tissue was removed from the distal femur and proximal tibia, care being taken to preserve knee ligaments and capsule. Bony members were then suspended in an alignment jig and polymethylmethacrylate (PMMA)-potted in tubular copper sleeves. These sleeves could be clamped within cylindrical adaptors in a specially-designed loading fixture (Fig. l), which, in turn, could be mounted between the ram and the load cell of a uniaxial material test system (MTS) testing machine. After potting and initial clamping in the loading fixture, the still-intact knee specimen was flexed to 40" [its position at the instant of maximal joint loading (l)] and the fat pad and anterior capsule proximal to the menisci removed, using sharp dissection. Next, the posterior capsule was resected, taking care not to cut posterior attachments of the menisci to the tibia. This allowed visualization of the joint
J Orthop Res, Vol. 9, No. 4, 1991
swivel pin,
1,
j tibia
femur
,_--flexion
pin
FIG. 1. Schematic diagram of specimen loading fixture. The swivel pin, passed anterioposteriorly midway between condylar midlines, assured reproducible, intercondylar load allocation.
line, anteriorly and posteriorly. Collateral ligaments were then detached from femurs and menisci, making it possible to rock each condyle's joint line slightly open, by manually applying either varus or valgus moments. A 0.5 mm steel marker pin (to be used as a reference point on subsequent Fuji film stains) was embedded in each condyle's articular cartilage, at a point just opposite the anterior edge of the corresponding meniscus. The loading fixture was then mounted in the MTS machine. With the knee locked in 40" of flexion, a small distraction load (50 N) was applied to facilitate insertion of a rectangular Fuji Pressensor film (medium pressure range) packet under each femoral condyle. Tensile load was then returned to 0 and the joint subjected to a 600 N (-3 times body weight) compressive load. Loading history was a linear upramp from 0 to 600 N in 0.1 sec, followed by a hold for 0.1 sec, and a linear downramp from 600 N to 0 in 0.1 sec. The joint was then again distracted by 50 N of tension to allow Fuji packet removal. (Four repeat trials were performed, with immediate visual examination of stains, to verify reproducibility of each condyle's Fuji contact pattern. One of these
CONTACT STRESS AROUND OSTEOCHONDRAL DEFECTS
four, subjectively identical stains was selected for quantitative analysis.) A point, approximating each intact surface's center of pressure, was then manually identified from the Fuji stain, and this pressure center point's distance from the corresponding marker pin's impression was measured with a ruler, marked in mm. The loading fixture was then dismounted from the MTS and the knee maximally flexed (to -130"), thus exposing marker pins and portions of the femoral condylar surfaces that had made contact with the tibia1 plateaus at 40" of knee flexion, a configuration coincident with the instant of peak loading during the stance phase of canine gait (12). The femoral condyle sites, approximating each intact surface's 40" center of pressure, were then identified (caliper measurement) relative to the imbedded marker pins. Osteochondral defects of 1 mm diameter were created at those central contact sites, using a trephine-like drill guide for first sharply cutting the articular cartilage around the defect rim, followed by guided drilling to a 4 mm depth, so as to remove interior cartilage and its underlying subchondral plate (Fig. 2). The loading fixture was then relocked at 40"of flexion and remounted in the MTS machine. The knee specimen was again distracted for Fuji packet insertion, then subjected to another 600 N load (again, with four repeated trials to ensure reproducibility). The loading fixture was again removed from the MTS, the knee again flexed to 130", the 1 mm defect concentrically enlarged to 2 mm (using a similar trephine/drilling procedure), and the knee again locked at 40" of flexion and returned to the MTS for contact stress mapping. The process was then repeated for defects, concentrically enlarged to 3 , 4 , 5 , 6 , and 7 mm. As a point of
561
reference, the area of a 7 mm diameter circular defect represents -60% of the cartilage contact area (series mean = 64.1 mm2, SD = 30.2), measured for normal, intact femoral condyle of the dog. Contact stress distributions were quantitated using a digital image analysis procedure, similar to that described by Singerman et a1 (25). Briefly, this involved backlighting the Fuji strips (Fig. 3A) in the view field of an Eikonix digital camera, with image capture in the form of a 512 X 512 pixel array of 8-bit, integer values (i.e., each pixel is assigned one of 256 possible grey scale levels). The Fuji images were digitized at a spatial resolution of -200 pixels/ mm2. Pixel arrays for captured images were converted from grey scale (Fig. 3B) to optical density (Fig. 3C) via Beer's law (27). Subsequent conversion from optical density to contact stress (Fig. 3D) was based on a series of calibration stains. Because defects were circular, systematic analysis was facilitated by transforming initial (Cartesian) contact stress data into polar arrays, with origins at respective defect centers. The defect center in each Fuji image was, first, manually identified by interactively superimposing cursor-defined circles to (subjectively) best fit the defect rim, apparent on the rendered Fuji image. Rays, emanating from the defect center, were then computationally constructed at 5" intervals (Fig. 3C). Contact stress values, at regular radial increments (of -0.07 mm) along each ray, were then read from corresponding pixel addresses in the original Cartesian array. Polar arrays of contact stress distribution were then compared across the specimen population as a function of defect size. Because perturbations of the flexion angle were noted to cause circumferential shifts of major features of the contact pattern (a phenomenon presumably occurring throughout the physiologic gait cycle), we elected to circumferentially average at each (pixel) radial increment (i.e., an average of the 72 individual ray values at a given radius), in order to make series-wide comparisons of the effect of defect size on contact stress concentrations. Circumferential averaging obviously tends to smooth the degree of contact stress concentration locally present at some sites around the defect rim, but provides a consistent means to catalog overall stress distribution changes. RESULTS
FIG. 2. Placement of typical defects (3 mm diameter) on the articular surface of medial and lateral femoral condyles.
For all seven defect sizes, in each of the 26 condyles studied, Pressensor stains demonstrated a
J
Orthop Res, Vol. 9, No. 4, 1991
562
1”. D . BROWN ET AL.
A B
D
C
FIG. 3. Raw Fuji stain photograph (A), digitized gray scale image (B), equivalent image after spatial filtering of crinkle artifact (C), and quantitative contact stress distribution in contour format (D), for a typical 3 mm defect. Manually identified regions of crinkle artifact were first delimited by cursor (dashed lines, B), and then filtered by resetting all enclosed pixels to a modified root mean square (RMS)intensity level (12th power, 12th root), noted empirically to minimize artifact contrast to neighboring, relatively homogeneously-staining regions.
tendency for contact stress concentration near at least a portion of the rim of the defect (Fig. 4). Several features of these images are noteworthy. First, despite careful alignment of a swivel pin, midway between respective condylar centerlines (Fig. l), substantial inequality of load-sharing between the two compartments nevertheless occurred. In all
J Orthop Res, Vol. 9, NO. 4, 1991
13 knees, integration of contact stress distributions revealed that load was preferentially transmitted through the lateral condyle. For the intact case, cartilage of the lateral condyle, on average, carried 64.1% (SD = 6.9%) of the total joint load, a percentage that differed insignificantly (p > 0.1 by paired Student t test) when “symmetrical” defects
563
CONTACT STRESS AROUND OSTEOCHONDRAL DEFECTS medial condyle
lateral condyle
(4
intact
(b)
2mm
4mm
(d)
6mm
increased with defect size and, for large defects (e.g., Fig. 4d) often involved development of radially tapering, lobe-like regions of excessive contact stress, which emanated from a relatively small portion (typically 40-70” of arc) of the defect rim. For the largest (7 mm) defects, rim geometry, apparent from the Pressensor, became decidedly noncircular in 15 of the 26 specimens. For the intact condyle, the series-average peak local contact stress was -6.2 MPa (SD = 2.63). This distribution was very uniform spatially (Fig. 5), the local value being within 1 MPa of “centerline” peak, out to a radius of slightly >2 mm, and within 2 MPa of peak, out to a radius of slightly >3 mm, radii corresponding to -50 and 75%, respectively, of the condyle’s entire cartilaginous width. A 1 mm diameter defect caused a small, irregularly demarcated central dip in the otherwise smooth contact stress profile. Peak contact stress (series average = 6.9 MPa, SD = 2.6) occurred at a radius of 1 mm, i.e., -0.5 mm outside the rim of the 1 mm diameter defect. On average, local contact stress decayed from peak by 1 MPa at an additional 0.5 mm radius, and by 2 MPa at an additional 1.2 mm. Intact-surface and 1 mm defect stress distributions were virtually indistinguishable beyond -2.5 mm from the defect center.
-
-IG. 4. Disturbance of contact patterns in a typical specinen, due to progressive enlargement of circular defects lo:ated centrally in the contact region. The respective panel lairs are (a) no defect, (b) 2 mm defect, (c) 4 mm defect, and :d)6 mm defect, for the medial and lateral condyles. Note the lresence of sharply delineated “lobes” of high contact stress ‘or the 6 mm defects and the fact that the contact stress jistributions around the rim of all defects are substantially ess uniform than are stresses in corresponding central reaions of intact condyles.
were present (e.g., 64.3%, SD = 6.7%, for bicondylar 7 mm defects). Preferential, lateral compartment load transmission was associated closely with lateral bias of the engaged contact area: an average of 63.7% (SD = 6.9%) of the total tibiofemoral contact area was on the lateral condyle for the intact surface case. Corresponding spatial, mean contact stresses were, on average, remarkably equal between compartments, the lateral being only 2.1% greater (SD = 0.12%) than the medial. While cartilage contact stress magnitude for intact condyles was relatively uniform within central regions of the contact zone, creation of wellcentered defects led to contact stress distributions that were substantially nonaxisymmetric about the defect centers. The degree of this nonaxisymmetry
R a d i u s (mm) 1
10
2 I
I
3 1
4
5 I
I
6
7
1
I
-
0
20
40
60
80
100
Radial Pixel Number
FIG. 5. Series-average(over 26 condyles) distributions of circumferential mean contact stress, as a function of radial distance from defect center. Heavy solid line represents the case of an intact condyle. Although radius at which contact stress peaks progressively moves outward (in accordance with the location of the defect lip), magnitude of peak contact stress is nearly independent of defect radius.
J Orthop Res, Vol. 9, No. 4, 1991
564
7: D . BROWN ET AL.
The most severe, series-averaged, local pressure elevations occurred for the 2 mm defect, peaking at a value of 8.1 MPa at a radius of 1.4 mm (ie., 0.4 mm outside the defect rim). Further increases in defect diameter led to lesser elevations in peak contact stress and to an ever-closer approach of the pressure peak to the defect rim. For 6 mm diameter defects, the largest size for which a reasonably circular rim shape was preserved, the series-average peak contact stress had fallen to 7.5 MPa, occurring -0.15 mm outside the defect rim. As defect diameter was progressively enlarged, there was a consistent tendency for mild elevation of contact stress at sites remote from the defect centerline and an associated tendency for additional contact area recruitment, peripherally. For the 6 and 7 mm defects, however, we noted a mild, statistically insignificant (p > 0.1 by paired t test) decrease in total contact area, from 64.1 (SD = 30.2) to 52.2 mmz (SD = 18.2). As a consequence, spatial mean contact stress rose from 4.68 MPa for the intact knees, to 5.75 MPa for the 7 mm defects. The very modest overall stress concentration factor, apparent from circumferential averaging, belies what was, in fact, an often striking tendency for excessive stress elevation along a small portion of the defect rim. Local stress concentrations, involving only a small portion of the defect rim, habitually occurred, especially for large defects. There was a positive correlation (Pearson r = 0.63) between defect diameter and degree of nonuniformity (i.e., circumferential standard deviation) of rim contact stress. For each defect size, if the 72 rim pixels (spaced at 5" increments around the defect perimeter) are first rank-ordered by contact stress magnitude for each specimen, and then averaged across specimens for each rank, the resulting distributions (Fig. 6) reflect the series-wide, peak local rim stresses and rim stress concentration nonuniformities, characteristic of each defect size. Although spatial mean and peak local contact stresses rose only mildly with defect enlargement, there was a dramatic increase in severity of the radial component of the contact stress gradient (i.e., the radialward variation of contact stress). Defects led to between a four- and sevenfold increase in severity of radial variation of contact stress (compared to the intact condyle case), with the site of maximal gradient usually lying -0.1-0.2 mm outside of the defect rim (Fig. 7). There was a consistent tendency to infer a gradual increase in the ra-
J Orthop Res, Vol. 9 , N O . 4, 1991
20
I
I
I
I
I
I
20
40
60
5 -
' \': 0. 0
L.-:
0
Circumferentiol Pixel Rank
FIG. 6. Series-average distributions of rank-ordered defect rim contact stress distributions. Compared to intact condyle (heavy solid lines), progressive defect enlargement leads to progressively more distribution nonuniformity and to substantially higher peak local contact stresses, although wholerim circumferential average contact stress increases only mildly.
dial component of the contact stress gradient over the last, several (supposedly noncontacting) pixels, lying within the defect, probably because of some combination of inward cartilage deformation at the Radius (mm) 2 I
1
20
1
0
20
3 I
40
4 I
5
I
60
6 I
80
7 I
100
Radial Pixel Number
FIG. 7. Series-average radial distributions of circumferential mean of radial component of contact stress gradient. Heavy solid line represents the case of an intact condyle. Note progressive outward movement of the site of peak contact stress gradient, in accordance with outward migration of the defect lip. Except for the relatively modest elevation of the peak stress gradient for the 1 mm defect case (1 1 MPalmm) versus the intact case (2.8 MPalmm), peak contact stress gradient was nearly independent of defect diameter.
CONTACT STRESS AROUND OSTEOCHONDRAL DEFECTS defect lip, interspecimen variability, and/or numerical differencing scheme used to evaluate contact stress gradient. DISCUSSION Our salient finding was the absence of severe, spatially-consistent elevation of contact stress around the rim of full-thickness, osteochondral defects, even when the defect spanned nearly the full width of the femoral condyle. Also, there was little direct relationship between defect diameter and degree of contact stress increase. While contact stress did tend to concentrate at selected locations just outside the defect rim, patterns observed were specimen-specific and joint-configurationdependent. This finding is similar to that of Huberti and Hayes (14), who observed irregular, specimenspecific contact stress elevations in visually normal cartilage surrounding chondromalacia patellae lesions. Another phenomenon, paralleling the Huberti-Hayes (14) and the Nelson et al. (22) models, was that de novo contact area recruitment kept approximate pace with the effective central contact area loss associated with defect enlargement. Occurrence of maximal contact stress slightly outside the defect rim, rather than at the rim per se, is consistent with contact stress distributions seen near step-off incongruities (4,13). Averaged over rim circumference and population, the highest stress concentration factor observed in the present series was only -1.3, and, interestingly, this occurred for relatively small (2 mm diameter) defects, which spanned only about one quarter of the condyle width. However, unlike the magnitude of contact stress, the radial component of the contact stress gradient showed striking changes from normal, even when “smoothed” by circumferential averaging. Undeniably, small cartilage defects often heal (sometimes with tissue remarkably similar to normal hyaline cartilage), whereas large defects do not (24). Explaining this phenomenon on the basis of progressive elevations of contact stress, with increasing defect size, is intuitively attractive, but data from the present study do not support that conclusion. However, the data are consistent with the view that elevated contact stress gradients, in normal cartilage near the defects, may somehow inhibit normal repair. We chose the canine femoral condyle as the prep-
565
aration for study because its relevant functional biomechanics have been thoroughly investigated (1,2), because its size (unlike that of rabbit knee) suffices for reliable contact stress mappings, and because its in vivo counterpart (22) is easy to work with and relatively inexpensive (unlike horse and monkey). In the interest of extrapolation to other models, however, we studied the full gamut of defect sizes, ranging from minimal involvement (1 mm diameter) to disruption of nearly the entire articular surface (7 mm diameter). In interpreting our data, several limitations need consideration. First, series-wide data were collected for only one loading position (40” flexion). We placed the femoral defect in a location that would fully contact the tibia1 surface when the knee was positioned, as at the instant of maximal loading. This had the effect of maximally loading the cartilage surrounding the defect. During normal function, however, only a portion of a similarlypositioned defect would contact the tibia, at most times. In the largest lesions, for which contact stress was often strongly asymmetric due to the defect rim’s proximity to the contact margin, we still observed trends of modest rim pressure elevation and strong pressure gradient elevation. This finding, plus the anecdotal observation that stress concentration sites shifted circumferentially around the defect rims when the present flexion angle was perturbed (Fig. €9, suggests that measured trends of contact stress change with defect size probably apply also for partially contacting defects. Second, we studied the knees at only one loading rate and, therefore, must assume (for the present) that similar behavior would occur for different loading histories. [The viscoelastic nature of cartilage is well documented (21) and the importance of reproducible fluid conditioning is well recognized, although aggregate load uptake at quasiphysiologic strain rates approximates that of a linear elastic solid.] Third, since the upper limit of contact stress gradient that can be reliably transduced by digitallyimaged Pressensor is only -10 MPdmm (Hale JE and Brown TD, unpublished observations), apparent circumferential averages in the range of 12-15 MPdmm probably underestimate the actual contact stress gradient. Limitations notwithstanding, these data suggest that elevated contact stresses on cartilage surrounding defects do not, per se, influence either healing of cartilage defects or subsequent degeneration. Based
J Orthop Res, Vol. 9, No. 4, 1991
566
11”. D . BROWN ET AL.
FIG. 8. Typical shifts of 4 mm defect contact pattern features, due to 5” perturbations of knee flexion angle from 25 to 55“. Note that the major effect is that the location of the contact region shifts approximately 100” circumferentially, whereas nominal size, shape, and distribution of staining intensity change only modestly.
on repair responses of experimental, intraarticular fractures in rabbits, Mitchell and Shepard (20) suggested that “compression of cartilage surfaces either creates a physical environment that allows certain chondrocytes to heal . . . or . . . prevents ingrowth of granulation tissue from the subchondral bone that might interfere with repair of hyaline cartilage.” While the superficial layer of articular cartilage can be a source of new cells for repairing (26), it seems unlikely that contact stresses determine whether or not a defect heals. Elevated contact stress gradients at an osteochondral defect rim-temporally sustained throughout the contact cycle-contrast to the relatively small (and temporally fluctuating) contact stress gradients characteristic of intact cartilage surface articulation. Also, strong gradients of contact stress are associated with the presence of elevated shear stress in the deep cartilage layers (13). Given the importance of pressure-driven, cyclic, interstitial fluid transport in normal cartilage (21,28), one could easily envision that high pressure gradients (and/or shear stresses) might interfere with the ability of chondrocytes, adjacent to defects, to proliferate and/or produce new matrix or interfere with maturation of granulation tissue arising from the marrow. The contact stress gradient for a 1 mm defect was
J Orthop Res, Vol. 9, No. 4, 1991
only about half as severe as that for a 2 mm defect, whereas, for defects >2 mm, the rim stress gradient was relatively independent of defect diameter. This suggests a “threshold” effect, possibly consistent with experimental observations of reasonable hyaline cartilage restoration (in defects
Effects of Osteochondral Defect Size on Cartilage Contact Stress Thomas D. Brown, David F. Pope, Joseph E. Hale, Joseph A. Buckwalter, and Richard A. Brand Biomechanics Laboratory, Department of Orthopaedic Surgery, The University of Iowa, Iowa City, / A , U.S.A.
Summary: Contact stress distributions were studied in vitro for 13 dog knees, with full-thickness osteochondral defects drilled in the weight-bearing area of both femoral condyles. Diameters of the circular defects were concentrically enlarged from 1 to 7 mm. Digitally-imaged Fuji film was used to record cartilage contact stress distribution on femoral condyles for each increment of defect diameter. All specimens showed at least some tendency for contact stress concentration at the rim of the defects. However, detailed distributions had large interspecimen variability and, within a given specimen, contact stress distributions became progressively more nonuniform around the defect rim as the diameter was enlarged. Averaged over the full series of 26 condyles, circumferential mean cartilage contact stress around the defect rim was only moderately higher (by 10-30%) than intact surface’s peak local contact stress [series average = 6.2 mega pascals (MPa)]. Maximal rim stress concentration occurred for 2 mm defects, there being a consistent trend toward mild rim stress decrease with further defect enlargement. Such modest contact stress elevations, per se, are probably insufficient to inhibit defect repair or to cause degeneration of surrounding cartilage. However, near the defect rim (for all diameters), the radial component of the gradient of contact stress (i.e., radialdirection variation of contact stress) was consistently elevated by an order of magnitude above that for intact, condyle articular cartilage. Key Words: Contact stress-Cartilage healing-Cartilage repair-Osteochondral defect.
While superficial articular cartilage defects possess little intrinsic capacity for repair (7,8,17,18), lesions extending through the subchondral bone show variable ability to repair (3,5,6,9-11,19,20, 22,24,26). Smaller lesions (i.e., up to 1-3 mm, depending upon species) tend to heal, sometimes with almost normal-appearing hyaline-like cartilage (24), while larger lesions (2-3 mm or more) tend to heal with fibrocartilage of variable quality (6,9,22). Clinically, the presence of a significant chondral defect
(i.e., clinically detected osteochondritis dissecans) is associated with a much higher incidence of osteoarthrosis than occurs in the general population, and clinically significant disease appears approximately a decade earlier than with idiopathic arthrosis (16). This clearly implies that some phenomenon, associated with defects, leads to arthrosis. Larger osteochondritic defects are associated with poorer clinical outcomes than are smaller ones (15), suggesting, perhaps, that some size-related mechanical phenomenon explains degeneration, as it might the quality of early repair. The reason (or cause) for variable repair is unclear. Based on substantial indirect evidence, Radin and Burr (23) suggested that while a mechanical
Received May 16, 1990; accepted September 21, 1990. Address correspondence and reprint requests to Dr. T. D. Brown, Department of Orthopaedic Surgery, The University of Iowa, Iowa City, IA 52242, U.S.A.
559
T. D . BROWN ET AL.
560
stimulus is essential for normal cartilage maintenance and repair, the magnitude of that stimulus is not large. They also suggested that reduction of local stress concentrations around defects favors repair. One could extend such reasoning and postulate that larger lesions heal poorly because their surrounding cartilage is subjected to deleteriously high contact stresses (i.e., contact force per unit surface area). In a recent in vivo canine study (22), we found that the highly compliant, fibrocartilaginous repair tissue, filling a large (6 mm) femoral condyle osteochondral defect, is incapable of moderating appreciable contact stress disruptions acutely present after defect creation. A related issue of clinical importance is the relationship between defect size and putative mechanically induced degeneration of surrounding cartilage. At present, however, the degree to which defect size, per se, causes local contact stress concentrations is not known. A study was, therefore, designed to test the hypothesis that cartilage surrounding large-diameter osteochondral defects experiences substantially elevated contact stress compared to that experienced around otherwise equivalent, small-diameter defects. MATERIALS AND METHODS Articular contact stress distributions were mapped around full-thickness osteochondral defects, created on both femoral condyles of 13 adult, mongrel dog knees. All specimens were excised, at necropsy, from 20-25 kg animals that had been normally active, prior to sacrifice, in the course of an unrelated acute study. Following hip disarticulation, all soft tissue was removed from the distal femur and proximal tibia, care being taken to preserve knee ligaments and capsule. Bony members were then suspended in an alignment jig and polymethylmethacrylate (PMMA)-potted in tubular copper sleeves. These sleeves could be clamped within cylindrical adaptors in a specially-designed loading fixture (Fig. l), which, in turn, could be mounted between the ram and the load cell of a uniaxial material test system (MTS) testing machine. After potting and initial clamping in the loading fixture, the still-intact knee specimen was flexed to 40" [its position at the instant of maximal joint loading (l)] and the fat pad and anterior capsule proximal to the menisci removed, using sharp dissection. Next, the posterior capsule was resected, taking care not to cut posterior attachments of the menisci to the tibia. This allowed visualization of the joint
J Orthop Res, Vol. 9, No. 4, 1991
swivel pin,
1,
j tibia
femur
,_--flexion
pin
FIG. 1. Schematic diagram of specimen loading fixture. The swivel pin, passed anterioposteriorly midway between condylar midlines, assured reproducible, intercondylar load allocation.
line, anteriorly and posteriorly. Collateral ligaments were then detached from femurs and menisci, making it possible to rock each condyle's joint line slightly open, by manually applying either varus or valgus moments. A 0.5 mm steel marker pin (to be used as a reference point on subsequent Fuji film stains) was embedded in each condyle's articular cartilage, at a point just opposite the anterior edge of the corresponding meniscus. The loading fixture was then mounted in the MTS machine. With the knee locked in 40" of flexion, a small distraction load (50 N) was applied to facilitate insertion of a rectangular Fuji Pressensor film (medium pressure range) packet under each femoral condyle. Tensile load was then returned to 0 and the joint subjected to a 600 N (-3 times body weight) compressive load. Loading history was a linear upramp from 0 to 600 N in 0.1 sec, followed by a hold for 0.1 sec, and a linear downramp from 600 N to 0 in 0.1 sec. The joint was then again distracted by 50 N of tension to allow Fuji packet removal. (Four repeat trials were performed, with immediate visual examination of stains, to verify reproducibility of each condyle's Fuji contact pattern. One of these
CONTACT STRESS AROUND OSTEOCHONDRAL DEFECTS
four, subjectively identical stains was selected for quantitative analysis.) A point, approximating each intact surface's center of pressure, was then manually identified from the Fuji stain, and this pressure center point's distance from the corresponding marker pin's impression was measured with a ruler, marked in mm. The loading fixture was then dismounted from the MTS and the knee maximally flexed (to -130"), thus exposing marker pins and portions of the femoral condylar surfaces that had made contact with the tibia1 plateaus at 40" of knee flexion, a configuration coincident with the instant of peak loading during the stance phase of canine gait (12). The femoral condyle sites, approximating each intact surface's 40" center of pressure, were then identified (caliper measurement) relative to the imbedded marker pins. Osteochondral defects of 1 mm diameter were created at those central contact sites, using a trephine-like drill guide for first sharply cutting the articular cartilage around the defect rim, followed by guided drilling to a 4 mm depth, so as to remove interior cartilage and its underlying subchondral plate (Fig. 2). The loading fixture was then relocked at 40"of flexion and remounted in the MTS machine. The knee specimen was again distracted for Fuji packet insertion, then subjected to another 600 N load (again, with four repeated trials to ensure reproducibility). The loading fixture was again removed from the MTS, the knee again flexed to 130", the 1 mm defect concentrically enlarged to 2 mm (using a similar trephine/drilling procedure), and the knee again locked at 40" of flexion and returned to the MTS for contact stress mapping. The process was then repeated for defects, concentrically enlarged to 3 , 4 , 5 , 6 , and 7 mm. As a point of
561
reference, the area of a 7 mm diameter circular defect represents -60% of the cartilage contact area (series mean = 64.1 mm2, SD = 30.2), measured for normal, intact femoral condyle of the dog. Contact stress distributions were quantitated using a digital image analysis procedure, similar to that described by Singerman et a1 (25). Briefly, this involved backlighting the Fuji strips (Fig. 3A) in the view field of an Eikonix digital camera, with image capture in the form of a 512 X 512 pixel array of 8-bit, integer values (i.e., each pixel is assigned one of 256 possible grey scale levels). The Fuji images were digitized at a spatial resolution of -200 pixels/ mm2. Pixel arrays for captured images were converted from grey scale (Fig. 3B) to optical density (Fig. 3C) via Beer's law (27). Subsequent conversion from optical density to contact stress (Fig. 3D) was based on a series of calibration stains. Because defects were circular, systematic analysis was facilitated by transforming initial (Cartesian) contact stress data into polar arrays, with origins at respective defect centers. The defect center in each Fuji image was, first, manually identified by interactively superimposing cursor-defined circles to (subjectively) best fit the defect rim, apparent on the rendered Fuji image. Rays, emanating from the defect center, were then computationally constructed at 5" intervals (Fig. 3C). Contact stress values, at regular radial increments (of -0.07 mm) along each ray, were then read from corresponding pixel addresses in the original Cartesian array. Polar arrays of contact stress distribution were then compared across the specimen population as a function of defect size. Because perturbations of the flexion angle were noted to cause circumferential shifts of major features of the contact pattern (a phenomenon presumably occurring throughout the physiologic gait cycle), we elected to circumferentially average at each (pixel) radial increment (i.e., an average of the 72 individual ray values at a given radius), in order to make series-wide comparisons of the effect of defect size on contact stress concentrations. Circumferential averaging obviously tends to smooth the degree of contact stress concentration locally present at some sites around the defect rim, but provides a consistent means to catalog overall stress distribution changes. RESULTS
FIG. 2. Placement of typical defects (3 mm diameter) on the articular surface of medial and lateral femoral condyles.
For all seven defect sizes, in each of the 26 condyles studied, Pressensor stains demonstrated a
J
Orthop Res, Vol. 9, No. 4, 1991
562
1”. D . BROWN ET AL.
A B
D
C
FIG. 3. Raw Fuji stain photograph (A), digitized gray scale image (B), equivalent image after spatial filtering of crinkle artifact (C), and quantitative contact stress distribution in contour format (D), for a typical 3 mm defect. Manually identified regions of crinkle artifact were first delimited by cursor (dashed lines, B), and then filtered by resetting all enclosed pixels to a modified root mean square (RMS)intensity level (12th power, 12th root), noted empirically to minimize artifact contrast to neighboring, relatively homogeneously-staining regions.
tendency for contact stress concentration near at least a portion of the rim of the defect (Fig. 4). Several features of these images are noteworthy. First, despite careful alignment of a swivel pin, midway between respective condylar centerlines (Fig. l), substantial inequality of load-sharing between the two compartments nevertheless occurred. In all
J Orthop Res, Vol. 9, NO. 4, 1991
13 knees, integration of contact stress distributions revealed that load was preferentially transmitted through the lateral condyle. For the intact case, cartilage of the lateral condyle, on average, carried 64.1% (SD = 6.9%) of the total joint load, a percentage that differed insignificantly (p > 0.1 by paired Student t test) when “symmetrical” defects
563
CONTACT STRESS AROUND OSTEOCHONDRAL DEFECTS medial condyle
lateral condyle
(4
intact
(b)
2mm
4mm
(d)
6mm
increased with defect size and, for large defects (e.g., Fig. 4d) often involved development of radially tapering, lobe-like regions of excessive contact stress, which emanated from a relatively small portion (typically 40-70” of arc) of the defect rim. For the largest (7 mm) defects, rim geometry, apparent from the Pressensor, became decidedly noncircular in 15 of the 26 specimens. For the intact condyle, the series-average peak local contact stress was -6.2 MPa (SD = 2.63). This distribution was very uniform spatially (Fig. 5), the local value being within 1 MPa of “centerline” peak, out to a radius of slightly >2 mm, and within 2 MPa of peak, out to a radius of slightly >3 mm, radii corresponding to -50 and 75%, respectively, of the condyle’s entire cartilaginous width. A 1 mm diameter defect caused a small, irregularly demarcated central dip in the otherwise smooth contact stress profile. Peak contact stress (series average = 6.9 MPa, SD = 2.6) occurred at a radius of 1 mm, i.e., -0.5 mm outside the rim of the 1 mm diameter defect. On average, local contact stress decayed from peak by 1 MPa at an additional 0.5 mm radius, and by 2 MPa at an additional 1.2 mm. Intact-surface and 1 mm defect stress distributions were virtually indistinguishable beyond -2.5 mm from the defect center.
-
-IG. 4. Disturbance of contact patterns in a typical specinen, due to progressive enlargement of circular defects lo:ated centrally in the contact region. The respective panel lairs are (a) no defect, (b) 2 mm defect, (c) 4 mm defect, and :d)6 mm defect, for the medial and lateral condyles. Note the lresence of sharply delineated “lobes” of high contact stress ‘or the 6 mm defects and the fact that the contact stress jistributions around the rim of all defects are substantially ess uniform than are stresses in corresponding central reaions of intact condyles.
were present (e.g., 64.3%, SD = 6.7%, for bicondylar 7 mm defects). Preferential, lateral compartment load transmission was associated closely with lateral bias of the engaged contact area: an average of 63.7% (SD = 6.9%) of the total tibiofemoral contact area was on the lateral condyle for the intact surface case. Corresponding spatial, mean contact stresses were, on average, remarkably equal between compartments, the lateral being only 2.1% greater (SD = 0.12%) than the medial. While cartilage contact stress magnitude for intact condyles was relatively uniform within central regions of the contact zone, creation of wellcentered defects led to contact stress distributions that were substantially nonaxisymmetric about the defect centers. The degree of this nonaxisymmetry
R a d i u s (mm) 1
10
2 I
I
3 1
4
5 I
I
6
7
1
I
-
0
20
40
60
80
100
Radial Pixel Number
FIG. 5. Series-average(over 26 condyles) distributions of circumferential mean contact stress, as a function of radial distance from defect center. Heavy solid line represents the case of an intact condyle. Although radius at which contact stress peaks progressively moves outward (in accordance with the location of the defect lip), magnitude of peak contact stress is nearly independent of defect radius.
J Orthop Res, Vol. 9, No. 4, 1991
564
7: D . BROWN ET AL.
The most severe, series-averaged, local pressure elevations occurred for the 2 mm defect, peaking at a value of 8.1 MPa at a radius of 1.4 mm (ie., 0.4 mm outside the defect rim). Further increases in defect diameter led to lesser elevations in peak contact stress and to an ever-closer approach of the pressure peak to the defect rim. For 6 mm diameter defects, the largest size for which a reasonably circular rim shape was preserved, the series-average peak contact stress had fallen to 7.5 MPa, occurring -0.15 mm outside the defect rim. As defect diameter was progressively enlarged, there was a consistent tendency for mild elevation of contact stress at sites remote from the defect centerline and an associated tendency for additional contact area recruitment, peripherally. For the 6 and 7 mm defects, however, we noted a mild, statistically insignificant (p > 0.1 by paired t test) decrease in total contact area, from 64.1 (SD = 30.2) to 52.2 mmz (SD = 18.2). As a consequence, spatial mean contact stress rose from 4.68 MPa for the intact knees, to 5.75 MPa for the 7 mm defects. The very modest overall stress concentration factor, apparent from circumferential averaging, belies what was, in fact, an often striking tendency for excessive stress elevation along a small portion of the defect rim. Local stress concentrations, involving only a small portion of the defect rim, habitually occurred, especially for large defects. There was a positive correlation (Pearson r = 0.63) between defect diameter and degree of nonuniformity (i.e., circumferential standard deviation) of rim contact stress. For each defect size, if the 72 rim pixels (spaced at 5" increments around the defect perimeter) are first rank-ordered by contact stress magnitude for each specimen, and then averaged across specimens for each rank, the resulting distributions (Fig. 6) reflect the series-wide, peak local rim stresses and rim stress concentration nonuniformities, characteristic of each defect size. Although spatial mean and peak local contact stresses rose only mildly with defect enlargement, there was a dramatic increase in severity of the radial component of the contact stress gradient (i.e., the radialward variation of contact stress). Defects led to between a four- and sevenfold increase in severity of radial variation of contact stress (compared to the intact condyle case), with the site of maximal gradient usually lying -0.1-0.2 mm outside of the defect rim (Fig. 7). There was a consistent tendency to infer a gradual increase in the ra-
J Orthop Res, Vol. 9 , N O . 4, 1991
20
I
I
I
I
I
I
20
40
60
5 -
' \': 0. 0
L.-:
0
Circumferentiol Pixel Rank
FIG. 6. Series-average distributions of rank-ordered defect rim contact stress distributions. Compared to intact condyle (heavy solid lines), progressive defect enlargement leads to progressively more distribution nonuniformity and to substantially higher peak local contact stresses, although wholerim circumferential average contact stress increases only mildly.
dial component of the contact stress gradient over the last, several (supposedly noncontacting) pixels, lying within the defect, probably because of some combination of inward cartilage deformation at the Radius (mm) 2 I
1
20
1
0
20
3 I
40
4 I
5
I
60
6 I
80
7 I
100
Radial Pixel Number
FIG. 7. Series-average radial distributions of circumferential mean of radial component of contact stress gradient. Heavy solid line represents the case of an intact condyle. Note progressive outward movement of the site of peak contact stress gradient, in accordance with outward migration of the defect lip. Except for the relatively modest elevation of the peak stress gradient for the 1 mm defect case (1 1 MPalmm) versus the intact case (2.8 MPalmm), peak contact stress gradient was nearly independent of defect diameter.
CONTACT STRESS AROUND OSTEOCHONDRAL DEFECTS defect lip, interspecimen variability, and/or numerical differencing scheme used to evaluate contact stress gradient. DISCUSSION Our salient finding was the absence of severe, spatially-consistent elevation of contact stress around the rim of full-thickness, osteochondral defects, even when the defect spanned nearly the full width of the femoral condyle. Also, there was little direct relationship between defect diameter and degree of contact stress increase. While contact stress did tend to concentrate at selected locations just outside the defect rim, patterns observed were specimen-specific and joint-configurationdependent. This finding is similar to that of Huberti and Hayes (14), who observed irregular, specimenspecific contact stress elevations in visually normal cartilage surrounding chondromalacia patellae lesions. Another phenomenon, paralleling the Huberti-Hayes (14) and the Nelson et al. (22) models, was that de novo contact area recruitment kept approximate pace with the effective central contact area loss associated with defect enlargement. Occurrence of maximal contact stress slightly outside the defect rim, rather than at the rim per se, is consistent with contact stress distributions seen near step-off incongruities (4,13). Averaged over rim circumference and population, the highest stress concentration factor observed in the present series was only -1.3, and, interestingly, this occurred for relatively small (2 mm diameter) defects, which spanned only about one quarter of the condyle width. However, unlike the magnitude of contact stress, the radial component of the contact stress gradient showed striking changes from normal, even when “smoothed” by circumferential averaging. Undeniably, small cartilage defects often heal (sometimes with tissue remarkably similar to normal hyaline cartilage), whereas large defects do not (24). Explaining this phenomenon on the basis of progressive elevations of contact stress, with increasing defect size, is intuitively attractive, but data from the present study do not support that conclusion. However, the data are consistent with the view that elevated contact stress gradients, in normal cartilage near the defects, may somehow inhibit normal repair. We chose the canine femoral condyle as the prep-
565
aration for study because its relevant functional biomechanics have been thoroughly investigated (1,2), because its size (unlike that of rabbit knee) suffices for reliable contact stress mappings, and because its in vivo counterpart (22) is easy to work with and relatively inexpensive (unlike horse and monkey). In the interest of extrapolation to other models, however, we studied the full gamut of defect sizes, ranging from minimal involvement (1 mm diameter) to disruption of nearly the entire articular surface (7 mm diameter). In interpreting our data, several limitations need consideration. First, series-wide data were collected for only one loading position (40” flexion). We placed the femoral defect in a location that would fully contact the tibia1 surface when the knee was positioned, as at the instant of maximal loading. This had the effect of maximally loading the cartilage surrounding the defect. During normal function, however, only a portion of a similarlypositioned defect would contact the tibia, at most times. In the largest lesions, for which contact stress was often strongly asymmetric due to the defect rim’s proximity to the contact margin, we still observed trends of modest rim pressure elevation and strong pressure gradient elevation. This finding, plus the anecdotal observation that stress concentration sites shifted circumferentially around the defect rims when the present flexion angle was perturbed (Fig. €9, suggests that measured trends of contact stress change with defect size probably apply also for partially contacting defects. Second, we studied the knees at only one loading rate and, therefore, must assume (for the present) that similar behavior would occur for different loading histories. [The viscoelastic nature of cartilage is well documented (21) and the importance of reproducible fluid conditioning is well recognized, although aggregate load uptake at quasiphysiologic strain rates approximates that of a linear elastic solid.] Third, since the upper limit of contact stress gradient that can be reliably transduced by digitallyimaged Pressensor is only -10 MPdmm (Hale JE and Brown TD, unpublished observations), apparent circumferential averages in the range of 12-15 MPdmm probably underestimate the actual contact stress gradient. Limitations notwithstanding, these data suggest that elevated contact stresses on cartilage surrounding defects do not, per se, influence either healing of cartilage defects or subsequent degeneration. Based
J Orthop Res, Vol. 9, No. 4, 1991
566
11”. D . BROWN ET AL.
FIG. 8. Typical shifts of 4 mm defect contact pattern features, due to 5” perturbations of knee flexion angle from 25 to 55“. Note that the major effect is that the location of the contact region shifts approximately 100” circumferentially, whereas nominal size, shape, and distribution of staining intensity change only modestly.
on repair responses of experimental, intraarticular fractures in rabbits, Mitchell and Shepard (20) suggested that “compression of cartilage surfaces either creates a physical environment that allows certain chondrocytes to heal . . . or . . . prevents ingrowth of granulation tissue from the subchondral bone that might interfere with repair of hyaline cartilage.” While the superficial layer of articular cartilage can be a source of new cells for repairing (26), it seems unlikely that contact stresses determine whether or not a defect heals. Elevated contact stress gradients at an osteochondral defect rim-temporally sustained throughout the contact cycle-contrast to the relatively small (and temporally fluctuating) contact stress gradients characteristic of intact cartilage surface articulation. Also, strong gradients of contact stress are associated with the presence of elevated shear stress in the deep cartilage layers (13). Given the importance of pressure-driven, cyclic, interstitial fluid transport in normal cartilage (21,28), one could easily envision that high pressure gradients (and/or shear stresses) might interfere with the ability of chondrocytes, adjacent to defects, to proliferate and/or produce new matrix or interfere with maturation of granulation tissue arising from the marrow. The contact stress gradient for a 1 mm defect was
J Orthop Res, Vol. 9, No. 4, 1991
only about half as severe as that for a 2 mm defect, whereas, for defects >2 mm, the rim stress gradient was relatively independent of defect diameter. This suggests a “threshold” effect, possibly consistent with experimental observations of reasonable hyaline cartilage restoration (in defects
