Introduction

Articular cartilage is a biologically active, relatively acellular complex tissue that provides near frictionless joint motion that is crucial for long-term function of diarthrodial joints [28]. The glycosaminoglycans (GAG) trapped within the collagen fibrils are negatively charged and generate swelling pressures, which carry the compressive load of the joint [21, 22]. Loss of GAG is one of the earliest events in cartilage degeneration. Histologically, the articular cartilage has a zonal organization where both the GAG content and structure of the collagen fibrils vary through the depth of the tissue [9, 16, 20]. In normal articular cartilage there are three major zones based on the orientation of collagen fibers: superficial, mid-, and deep layer [29], which can be distinguished on MR images [27]. In the superficial zone, the collagen fibrils are arranged parallel to the articular surface, whereas in the deeper zone, they are laid out perpendicular to the underlying bone [1, 16]. In addition to the variation in matrix structure, the volume density of chondrocytes and biosynthetic activity varies with the depth of the cartilage [17, 34, 43]. Many studies have shown that mechanical loading of articular cartilage affects the metabolism of chondrocytes and its biochemical composition [1, 32, 37]. Both animal and human studies have shown that the GAG content is higher in cartilage that is habitually loaded [19, 33] or has a higher level of activity [39], whereas immobilization results in a reversible decrease in cartilage PG content [18, 30].

Compositional MRI techniques such as delayed gadolinium-enhanced MRI of cartilage (dGEMRIC), T2, T1rho, T2*, gagCEST, and sodium imaging all attempt to measure the biochemical composition of cartilage [12]. T2, T1rho, and dGEMRIC have been studied extensively in the knee, whereas most studies in the hip, to date, used dGEMRIC. In dGEMRIC, a negatively charged contrast agent, gadolinium, is injected intravenously. Because of its negative charge, gadolinium will be repelled by tissues that also have negative charges such as cartilage with high GAG content. However, gadolinium will accumulate in areas with low charge meaning that lower GAG concentration leads to higher contrast agent distribution. This can be measured using MRI [2–4, 6]. Lattanzi et al [24] reported accuracy of 58%, sensitivity of 52%, and specificity of 67% in detecting cartilage damage in femoroacetabular impingement using dGEMRIC.

In acetabular dysplasia (AD) [11], an underdeveloped or shallow acetabulum leads to smaller contact areas and higher contact pressures in the hip compared with healthy subjects [25, 26, 40]. Periacetabular osteotomy (PAO), a hip-preserving procedure, is performed to restore normal anatomy and has been shown to evenly distribute stresses through the hip during weightbearing [45]. By simulating a reorientation pelvic osteotomy, it has been demonstrated that up to a 50% decrease in contact pressure can be achieved [15]. Given the importance of the weightbearing status on the biology of chondrocytes [23, 31] and the heterogeneous structure of cartilage [7], our study investigated whether the mechanical modulation of the dysplastic joint with a PAO affects the biochemical composition of articular cartilage and whether this effect varies with the depth of cartilage. The superficial cartilage layer has been shown to have a low charge density, low compressive stiffness [8], and to be less protected by inhibitors to degradation [13], which suggest that this region may be more vulnerable. A better understanding of the biologic changes in articular cartilage with a pelvic osteotomy may help us better predict the long-term health of the joint.

Therefore, we wished to (1) investigate if the superficial zone behaves differently than the deep zone after PAO given that the two layers are histologically and biologically different; and (2) if the superficial and deep zones behave the same way in different areas of the joint.

Materials and Methods

This is a prospective single-group longitudinal study. Institutional review board approval was obtained before the start of enrollment.

This cohort was used in a prior paper analyzing the clinical outcomes and dGEMRIC indices at the acetabular and femoral cartilage after PAO [14]. The present article is looking specifically at the effect of the PAO in the superficial and deeper cartilage zones.

Study Population

Inclusion criteria were the following: the diagnosis of acetabular dysplasia with a lateral center-edge angle smaller than 20° and the presence of hip-related pain. Less than 90° flexion, signs of advanced osteoarthritis (Tönnis Grade II or more [25]) as well as neuromuscular disorders or chromosomal disorders were exclusion criteria. Also the presence of an incongruous hip (“fair” or “poor” in the Yasunaga classification [44]) as seen on von Rosen view [41] was an exclusion criterion.

A total of 136 patients were screened during the study period and 53 patients were enrolled in the study. Of 53 patients, 16 patients did not return for their 1-year dGEMRIC scans; therefore, 37 patients (37 hips) constituted the study cohort. Patients were scheduled for visits preoperatively (baseline), at 6 months, 12 months, and 24 months after surgery. Of the 37 patients, 28 had 2-year followup with measurements at all three visits, whereas the other nine patients had only 1 year of followup.

Of the 37 patients, 34 were female and three were male. Thirty-four patients were white, one was black, and two were Asian. The mean age (± SD) at the time of the preoperative MRI was 26 ± 9 years (range, 13–46 years). Twenty-seven (73%) were right hips and 10 were left hips. Twenty-eight of the 37 patients had 2 years of followup with measurements at all three visits, whereas the other nine patients had only 1 year of followup. There were no differences in preoperative or 1-year dGEMRIC indices between subjects with 1 year of followup and subjects with 2 years of followup (Table 1). There was also no difference in the change in mean dGEMRIC T1 index from preoperative to 1-year followup across the differing followup groups.

Table 1 Summary of dGEMRIC T1 values (msec) separated by subjects with complete followup and those who missed their 2-year followup visit
Full size table

Magnetic Resonance Imaging

In all cases a standardized protocol was used for dGEMRIC [6]. dGEMRIC was performed on a single 1.5-Tesla system (Magnetom Avanto; Siemens Healthcare, Erlangen, Germany) with a flexible surface coil. After intravenous gadolinium injection (0.2 mM/kg gadolinium-DTPA2-; Magnevist®; Berlex/Bayer HealthCare Pharmaceuticals Inc, Wayne, NJ, USA), patients walked for a minimum of 15 minutes to accelerate contrast diffusion into the hip. The MR scans started approximately 22 minutes after contrast administration and the isotropic T1 mapping sequence at a mean of 48 minutes after Gd-DTPA2- (SD, 12 minutes; range, 30–70 minutes).

A three-dimensional isotropic gradient echo dual-flip angle T1 mapping sequence [10, 38] was used to obtain the T1 map with the following parameters: repetition time 15 ms, echo time 4.68 ms, flip angles of 5° and 28°, a matrix size of 192 × 192, 160 × 160-mm field of view, 96 slices, isotropic voxel size 0.83 mm, and acquisition time of 6 minutes 51 seconds. An isotropic True-FISP sequence was used for morphological images with the following parameters: repetition time 12.57 ms, echo time 5.48 ms, a matrix size of 256 × 256, 160 × 160-mm field of view, 144-slice slab, and an isotropic voxel size of 0.63 mm. The scan time was 7 minutes 47 seconds.

Both the isotropic data sets of the T1 mapping and True-FISP sequence were acquired in the oblique axial plane and used for radial reconstruction and evaluation on a Leonardo workstation (Siemens Healthcare, Erlangen, Germany). As described in other studies [5, 10], we reconstructed five acetabular radial reformats in 30° steps (Fig. 1) based on anatomic landmarks of the acetabulum. This allows us to look at the same regions even after a rotational pelvic osteotomy. The reformats allowed image evaluation at the following positions: anterior (A), anterosuperior (AS), superoanterior (SA), superior (S), and superoposterior (SP). Their slice thickness was 3.0 mm for the T1 mapping and 1.2 mm for the True-FISP data set.

Fig. 1
figure 1

Five radial reformats around the acetabular opening are generated from three-dimensional MRI: Anterior (1), AS (2), SA (3), S (4), SP (5) planes around the acetabulum are generated by first identifying the transverse acetabular ligament, which is used as the reference landmark. The superior plane goes through the middle of the transverse ligament and the remaining planes rotate in 30º steps around the acetabular opening axis.

Full size image

dGEMRIC Evaluation

The region of interest (ROI) evaluation was performed manually by a trained reader (AMH) on anonymized T1 data sets. The superficial zone of the femoral and acetabular cartilage was combined to one ROI and defined as the superficial zone. A ROI of equal size was then selected for the subsequent deeper acetabular cartilage and defined as the deeper zone (Fig. 2).

Fig. 2
figure 2

Schematic illustration shows the dGEMRIC ROI evaluation.

Full size image

The ROI was selected, in direct comparison to morphological images, in the weightbearing area, defined by the acetabular rim as the peripheral border and the acetabular fossa as the central border.

Interobserver variability for this type of evaluation has been shown to be excellent with intraclass correlation coefficients of 0.92 [5, 10]. Counting all five radial reformats, we evaluated 370 ROIs preoperatively, 370 ROIs at 12-month, and 280 ROIs at 24-month followup.

Statistical Analysis

To evaluate the differences in the superficial and deeper cartilage zones, dGEMRIC T1 indices from the five radial planes in both the superficial and deep zones from each patient were analyzed. Changes in dGEMRIC T1 measurements were evaluated at three time points (preoperative, 1-year followup, and 2-year followup) during the study using mixed model analysis. For each time point, each subject had a possible 10 measurements (five radial plane measurements in the superficial zone and five radial plane measurements in the deep zone). Mixed model analysis was used to describe the change in dGEMRIC T1 values while accounting for within-subject correlations as a result of repeated measures on each subject. All data were analyzed assuming an autoregressive correlation structure for the covariate matrix and under the assumption that any missing data were missing at random. Reported SDs are between-subject SDs of the total dGEMRIC values that were obtained by averaging over all 10 regions measured within each patient. To assess whether loss of followup was dependent on baseline or 1-year measurement, dGEMRIC T1 indices were compared between subjects who were missing their second-year followup visit and subjects with complete 2-year followup information using intercept-only mixed models (Table 1). Secondary analysis was conducted to assess variation in dGEMRIC T1 measurement across the radial plane of the acetabulum. Multivariate linear mixed models were used to analyze the change in dGEMRIC T1 index between the superficial and deep zones across the five radial planes (A, AS, SA, S, and SP) at each of the three time points. All tests were two-sided and p values < 0.05 were considered significant. Analyses were performed using SAS software Version 9.3 (SAS Institute Inc, SAS, Cary, NC, USA).

Results

Changes in dGEMRIC Index: Superficial versus Deep Layers

GAG content, as measured by dGEMRIC, was lower in both the superficial and deep zones 1 year after surgery and recovered slightly by Year 2. Greater fluctuation in dGEMRIC index was seen in the superficial zone (Table 2), falling from a mean (± SD) of 480 (± 137 msec) preoperatively to 409 (± 119 msec) (p < 0.001) at Year 1 and recovering to 451 (± 115 msec) at Year 2 (p < 0.001). In the deep zone, the dGEMRIC index fell from 574 (± 142 msec) preoperatively to 526 (± 140 msec) at Year 1 (p < 0.001) and recovered to 537 (± 121 msec) at Year 2 (p = 0.008). However, over the entire 2-year span of the study, dGEMRIC index decreased in the superficial zone by an average of 22 msec (95% confidence interval [CI], −40 to −4; p = 0.02), whereas in the deep zone, dGEMRIC index decreased by an average of 36 msec (95% CI, −56 to −15; p < 0.001). An example of a patient scan is shown (Fig. 3). The superficial zone could be seen with a substantial decrease at 1-year postoperative scan compared with the preoperative scan.

Table 2 Summary of dGEMRIC T1 values and mixed model estimated slopes for all patients (n = 37) representing the change in T1 (msec) for each time point by superficial and deep zone
Full size table
Fig. 3
figure 3

A 40-year-old woman underwent right PAO for symptomatic dysplasia. Postoperative radiographs show good correction with no progression of osteoarthritis in 2 years. The preoperative dGEMRIC scan shows the superficial zone with lower dGEMRIC index, as indicated by the arrow. The 1-year postoperative scan shows the same region (arrow) with a lower dGEMRIC index in the superficial zone. The 2-year scan shows perhaps some recovery of the dGEMRIC index in the superficial zone.

Full size image

Variation in dGEMRIC Index: Across the Radial Planes

In the differing acetabular planes of the hip, the superior region superficial layer dGEMRIC index had the most variation after surgery. dGEMRIC index varied across the radial planes in both the superficial (p < 0.001) and deep acetabular zones (p < 0.001) for each of the three time points in the study (Fig. 4). In the superficial zone, decreases were seen in dGEMRIC index at the AS plane (p = 0.008), the SA plane (p < 0.001), and the S plane (p < 0.001) from preoperative to 1-year followup and no change was detected at the A (p = 0.28) or SP (p = 0.45) planes. In the deep zone, there was a decrease in dGEMRIC measurement from preoperative to 1-year followup in the SP plane (p = 0.03) and in the S plane (p = 0.01), but there were no changes in the A (p = 0.20), SA (p = 0.06), or AS (p = 0.60) planes. From 1-year to 2-year followup, the only detected change in dGEMRIC index was an increase in the superficial zone at the S plane (p = 0.006). No changes were detected in any of the other radial planes in the superficial zone (p = 0.26–0.80). There were no differences observed in the deep acetabular zone across any of the radial planes (p = 0.43–0.86) from 1-year to 2-year followup. From preoperative to 2 years, an overall decrease in dGEMRIC was detected in the superficial zone only at the S (p = 0.003) and SA (p = 0.002) planes (Table 3). In the deep zone, a decrease from preoperative to 2 years was found only at the SP (p = 0.009) and S (p = 0.008) planes.

Fig. 4
figure 4

Average T1 values (dGEMRIC index) at radial planes in the superficial and deep zones at all visits are shown.

Full size image
Table 3 Summary of dGEMRIC T1 values (msec) by radial plane at the superficial and deep zones
Full size table

Discussion

The long-term goal of PAO for acetabular dysplasia is to preserve hip function. It is presumed that normalization of the abnormal mechanics with improved coverage of the femoral head will alter the initial normal biologic response (hypertrophic) and then abnormal response (catabolic) of the articular cartilage to the abnormally increased mechanical load. This type of response is certainly seen in tissue culture and animal studies of cartilage response to mechanical loading. Our initial analysis of this cohort [14] supports this hypothesis by demonstrating that in areas with increased mechanical load, the charge density of the acetabular cartilage as measured by dGEMRIC decreased to normal levels after osteotomy. This interpretation of our findings seems quite consistent with what we know about cartilage mechanoregulation; however, because we visually inspect the data, we noticed that the changes in appearance of the superficial layer seemed different from the deep layer. Attributable to the fact that in vitro studies have shown that there are depth-related differences in mechanical behavior [8, 42] and biochemical properties [17, 36] in response to compressive loading of articular cartilage, we have decided to analyze our data further. Therefore, our goals were twofold: (1) to compare the response of the superficial layer cartilage to that of the deep layer after PAO in dysplastic hips; and (2) to assess whether these different responses were confined to the acetabular region that sees alteration in mechanical load.

The present study has some limitations. First, because we combined the superficial zones of the acetabular and femoral cartilage in one ROI as a result of the thin articular cartilage layer, we included the potential space between the two cartilage layers, where the synovial fluid is considered to have the potential to affect the dGEMRIC index. To overcome this limitation we evaluated all MRI reformats for the presence of synovial fluid. However, no synovial fluid layer could be detected on a single morphological True-FISP reformat, most likely as a result of the extreme thinness of this layer. Furthermore, it has been reported that the protein composition of the superficial layer is almost indistinguishable from that of synovial fluid [9]; therefore, we considered that our approach is reasonable for evaluating the superficial zone. Second, the lack of a control group (natural history group) is a limitation. Finally, we have performed inter- and intraobserver reliability studies for similar analysis in other studies; however, we did not do so specifically for this study.

Chen et al [8] investigated the relationship between compressive properties of the different cartilage layers and their fixed charge density, which reflects the amount of GAG. They found lower fixed charge density in the region near the articular surface, which was associated with lower compression properties. This can be confirmed by our results, demonstrating lower dGEMRIC values in the superficial compared with the deeper zones at all visits. Furthermore, similar results have also been found for bovine knee cartilage [34, 35], suggesting that the extent of the depth dependence of compressive properties with the lowest values near the articular surface may be a fundamental property for tissue with normal biomechanical function.

Over time, we found a decrease in mean dGEMRIC index in both superficial and deep zones at 1-year followup, which appears to recover somewhat at the 2-year followup. On closer inspection, the biochemical changes in cartilage after PAO were greater in the superficial than in the deeper cartilage zone, suggesting a higher response of the superficial zone compared with the deeper zone to changes in biomechanics or the possible nonbiomechical factors including the inflammatory process in the joint after surgery. This confirms the findings of several other authors, indicating depth-related properties and metabolic specialization of the resident chondrocytes leading to characteristic zonal variations in cartilage adaption to mechanical loading [9, 36, 43]. In addition, the unique depth-dependent features and high vulnerability of the superficial layer have been demonstrated in a study investigating the effect of inflammatory cytokines in different cartilage layers [13]. They reported at least a 10-fold greater concentration of interleukin-1α for a similar inhibition of PG synthesis in cells from the deeper layer compared with the superficial layers.

When the variation in dGEMRIC indices were investigated further by looking across the five radial planes, the most responsive region was seen mostly in the superficial layer of the superior part of the acetabulum, whereas in the deep zone, no difference across the radial planes could be detected from 1- to 2-year followup. We hypothesized that postoperative protected weightbearing, inflammation, and overall normalization of mechanical load within the joint will affect the biosynthetic activity of chondrocytes [32, 42] and thus leads to characteristic changes in the dGEMRIC index at followup visits. Given that the effect of PAO was most pronounced at the superior aspect of the joint, where the mechanical forces are the greatest in patients with AD, indicates that our results may be truly reflecting the cartilage adaption to normalization in mechanical loading [45].

Our study has shown that the superficial zone does have a consistently lower charge density than the deep zone, which is the normal variation seen in normal cartilage. In addition, it appears that the superficial zone is more sensitive to alteration in biomechanics of the hip after PAO compared with the deep zones. In both the deep and superficial zones, there was an initial decrease seen at 1 year, which partially recovered at 2 years. The magnitude of change was bigger in the superficial zone. In both the superficial and deep zones, most of the change in dGEMRIC index was confined to the acetabular region near the superior part of the joint, which is where the alteration in mechanics would be largest. Most of the change in dGEMRIC index seems consistent with the alteration in mechanics; however, the large drop in dGEMRIC index and then recovery at 2 years, especially in the superficial zone of the superior part of the acetabulum, suggests that there may be additional effects such as postoperative inflammation. Perhaps this may explain the clinical observation that some hips may have an acceleration of degeneration with surgical intervention.