Introduction

Osteoarthritis (OA) causes pain and disability for the aging population and has considerable impact on healthcare costs [9]. Hip and knee arthroplasties have become two of the most common surgical procedures performed today, and the numbers of hip and knee arthroplasties performed are expected to double from 2000 to 2020 [22]. How OA develops is not well understood, but it is now accepted that biomechanical abnormalities in the joint that result in repetitive joint injury can initiate and accelerate the disease process. Such is the case with developmental dysplasia of the hip (DDH) and femoroacetabular impingement (FAI), with their characteristic abnormal morphologies of the hips. By some estimates, 20% to 40% of hips requiring arthroplasties may be due to secondary OA from DDH [3, 26, 64]. The smaller area of anterolateral acetabular coverage in DDH increases mechanical stress in the cartilage, which results in cartilage damage [28]. FAI with variable anatomic abnormalities of the acetabulum and/or femoral head is also a more recently recognized contributor to OA by mechanical impingement leading to direct stress to the articular cartilage or via damage to the labrum [5, 71].

Traditionally, the diagnosis of OA is made using radiographs, which indirectly assess cartilage damage by measuring apparent joint space loss (distance between acetabular and femoral subchondral plates) or bony reaction to the arthritis. However, radiographs cannot detect early cartilage injuries [21]. Conventional MRI allows direct visualization of the morphology of cartilage, including fissuring or partial- to full-thickness cartilage loss as a result of OA. With higher field strength magnets, improved visualization of morphologic changes in cartilage is possible, but detection of compositional changes, which usually precede morphologic changes, is not possible with conventional MRI.

Mature hyaline cartilage is composed of relatively few chondrocytes (2%) and the rest is composed of extracellular matrix comprised of water (70%) and macromolecules, predominately Type II collagen fibers and proteoglycan molecules. Early OA is characterized by loss of proteoglycans, which is followed by loss of collagen and changes in collagen orientation [74]. These biochemical properties in cartilage can deteriorate with mechanical overload without any radiographic changes. To improve our understanding of the natural history of OA development and to detect more sensitive and rapid outcome measures of pharmacologic and surgical interventions, recent studies have focused on using proteoglycan, collagen, and/or water content in hyaline cartilage as biomarkers for quantifying early OA and its progression.

Several compositional technique MRIs are available: T2 mapping, T1ρ, diffusion-weighted imaging, sodium MRI, and delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) (Appendix 1). These MRI techniques have the potential to become valid imaging biomarkers that may affect the management of clinical therapies. At present, the best validated and commonly utilized technique for the hip is dGEMRIC. dGEMRIC utilizes a negatively charged gadolinium-based contrast agent, which acts as a probe to measure the negative charge density of cartilage. It is a specific measure of glycosaminoglycan content of cartilage. dGEMRIC has been validated against histology in vitro and ex vivo and clinically in hips and knees [4, 10, 15, 31, 42, 51, 61, 68]. The dGEMRIC index reportedly correlates with hip pain and degree of hip dysplasia [31]. Similar findings were also seen in patients with FAI, where dGEMRIC correlated with pain and alpha angle [29]. Additionally, dGEMRIC mapping of hips with history of slipped capital femoral epiphysis demonstrated differences between hips with normal and severe offsets [79].

Bernese periacetabular osteotomy (PAO) is effective for mild to severe hip dysplasia in relieving hip pain and improving joint biomechanics. This osteotomy may slow the progression of secondary OA and delay the need for salvage surgery, arthrodesis, or arthroplasty [11, 16, 17, 27, 50, 55, 57, 62, 65, 73]. In general, early reconstructive osteotomies performed on younger patients who have little radiographic evidence of OA by Tönnis grades are regarded the best candidates for this procedure [16, 33, 41, 47, 50, 73]. We expect hips that fail prematurely after osteotomy would have more evidence of OA on radiographs. In our initial study [15], we demonstrated the coronal dGEMRIC index best predicted pain improvement and lack of radiographic disease progression and avoidance of arthroplasty after PAO. We initially utilized just the coronal MRI cuts in our dGEMRIC assessment; however, we now recognize regional variations in the mechanical stress in the hip should lead to nonuniform degenerative changes. Since the damage is most severe in the anterior part of the joint, selective sagittal cuts may be a more sensitive predictor of outcome after PAO. Based on previous dGEMRIC studies, the coronal and sagittal slice dGEMRIC values would correlate to one another; however, we predict the anterior dGEMRIC index, which would have the most amount of damage, would best predict premature joint failure after osteotomy.

We therefore asked whether the regional variations in biochemical composition in cartilage in dysplastic hips predict failure based on pain and function, progression of apparent joint space loss, and subsequent THA after PAOs.

Patients and Methods

We retrospectively analyzed a subset of a cohort from a previously published prospective study [15]. The initial cohort consisted of 207 hips from 171 patients with DDH who underwent Bernese PAO procedures from October 1999 to November 2002. Inclusion criteria for the patients selected for this study were as follows. All patients were offered Bernese PAO, a redirectional PAO, for activity-related hip pain and radiographic evidence of acetabular dysplasia [12, 23]. Patients with only primary acetabular osteotomies were included. However, if patients previously had open reduction and intertrochanteric osteotomies before or at the time of the primary acetabular osteotomy, they were included. At the time of the surgery, all patients in the cohort were at least near skeletally maturity with closed triradiate cartilage. They needed to have hip flexion of greater than 90° and radiographic evidence of acetabular dysplasia, defined as a lateral center-edge angle (LCE) of less than 20°. Patients were excluded if they had an incongruous hip as seen on abduction/internal rotation pelvic radiographs or if there was evidence of arthritis of greater than Tönnis Grade 3 [70] on radiographs. Any hip dysplasia caused by neuromuscular disorder or skeletal dysplasia was also excluded. Of 52 hips from 47 patients who met the inclusion criteria for the previous study, only the hips with both preoperative coronal and sagittal dGEMRIC scans before undergoing Bernese PAO were included in this study [15]. The cohort for this study consisted of 43 hips from 41 patients (36 female, five male). Clinical and radiographic images were obtained before and after the PAO. The minimum followup was 24 months (mean, 32 months; range, 24–46 months). No patients were lost to followup. No patients were recalled specifically for this study; all data were obtained from medical records and radiographs. Our institutional review board approved this study, and the patients or their parents provided informed consent for the surgery.

Preoperative clinical assessments were made using WOMAC index [6] questionnaires. Pain score for each hip was the sum of the scores for the five questions in the pain assessment section.

Preoperatively, all patients had standing AP and false-profile radiographs obtained of the affected joint and a supine AP view of the pelvis with the hip maximally abducted and internally rotated. During postoperative followup scheduled for at least 6 months after the surgery, only the standing AP pelvis and false-profile views of the affected joint were obtained. On the AP pelvic film, two of us (SDK, RJ) measured the joint space width (JSW) as the minimum radial distance between the subchondral bone of the acetabulum and the femoral head in the weightbearing area [14]. The LCE angle was measured as the angle formed by a vertical line drawn through the center of the femoral head and a line drawn along the edge of the acetabulum on the AP pelvic radiograph [77]. Tönnis grade and Tönnis angle were measured on AP pelvic radiographs [70]. Joint subluxation was defined as a break in Shenton’s line of 5 mm or more [13]. Anterior center-edge (ACE) angle was measured using the false-profile view [34]. The von Rosen view (maximal abduction, internal rotation) was used to assess joint congruency preoperatively and postoperatively with AP pelvic radiographs using the classification of Yasunaga et al. [78]. The morphologic measurements are associated with high inter- and intrareader reliability with intraclass correlation coefficients ranging from 0.9 to 0.97 [18]. The JSW measurements are associated with inter- and intrareader intraclass coefficients of 0.68 to 0.86 [7].

The dGEMRIC scans were performed as described previously, but the current study included analysis of the multislice fast spin echo (FSE) sagittal slices with saturation recovery technique. The dGEMRIC scans were performed using a GE 1.5-T clinical scanner (General Electric Corp, Milwaukee, WI, USA). A double dose (0.4 mL/kg) of Magnevist® (gadolinium diethylene triamine pentaacetic acid [Gd-DTPA2−]; Bayer HealthCare, Wayne, NJ, USA) was injected intravenously. The patients were required to walk for 30 minutes before the scan to ensure full penetration of the gadolinium into the articular cartilage. The following imaging parameters were used: echo time of 14 milliseconds and repetition times of 300, 500, 750, 1000, 1500, and 2000 milliseconds [15, 31]. T1 value was calculated using custom Matlab® software (Mathworks, Natick, MA, USA) to perform a pixel-by-pixel fit of the image intensity of the six FSE image data to a rising exponential equation. Four coronal slices with 4-mm slice thickness and 1-mm spacing were obtained. Two to three sagittal slices were obtained with 3-mm thickness, 1-mm spacing, and in-plane resolution of ~ 0.7 mm. The T1 values of the femoral and acetabular cartilages were averaged to calculate the dGEMRIC index. On the coronal slices, the coronal dGEMRIC index of the weightbearing cartilage was calculated as the average T1 value of the femoral and acetabular cartilages from the lateral edge of the bony acetabulum to the edge of the acetabular fossa. On the sagittal slices, the anterior sagittal dGEMRIC index was calculated as the average T1 value of the femoral and acetabular cartilages from the anterior edge of the bony acetabulum to a vertical line going through the center of the head. The posterior sagittal dGEMRIC index was the average T1 value of the femoral and acetabular cartilages from the vertical line going through the center of the femoral head to an equal arc used to calculate the anterior sagittal dGEMRIC index. The sagittal dGEMRIC index was the average of the anterior and posterior sagittal dGEMRIC indexes.

Our routine clinical followups were performed at 1, 2, 3, 6, and 12 months and then yearly after the osteotomy. At each visit, an AP pelvic radiograph was obtained. Usually at either the 3- or 6-month visit, false-profile radiographs were obtained to assess the acetabular correction. At the 6-month and yearly visits, WOMAC score and detailed clinical hip ROM and examination were obtained. For longer-term followups, nine patients were interviewed over the telephone by trained research assistants or filled out the WOMAC questionnaires sent to them by mail. Additionally, some plain radiographs were obtained locally and sent to us for analysis.

The surgical intervention was deemed a failure if the patient required a THA during the followup period, had persistent postoperative WOMAC pain scores of greater than 10 points, or had minimal JSWs decreased to less than 3 mm. Among the 43 hips, five hips eventually required a THA during the followup period, and one hip had a decrease in JSW to less than 3 mm. The final study cohort consisted of two groups, one with 37 hips that underwent a successful pelvic osteotomy and the other with six hips for which the operation was deemed a failure (Table 1). All six patients whose osteotomy failed were women with a mean (± SD) age of 34 ± 7 years, compared to the patient population who had not failed by the criteria above (32 female, five male) with a mean age of 28 ± 11 years (p = 0.21) (Table 1). The mean differences between the PAO satisfactory and PAO failure groups were 5.9 years (95% CI, 2.0–13.0 years) for age, 5.4 points (95% CI, 2.0–9.0 points) for WOMAC pain score, −1.9 mm (95% CI, −1.0 to −3.0 mm) for JSW, −201 milliseconds (95% CI, −116 to −286 milliseconds) for coronal dGEMRIC index, and −261 milliseconds (95% CI, −154 to −367 milliseconds) for sagittal weightbearing dGEMRIC index.

Table 1 Demographic and preoperative radiographic data stratified according to results of PAO
Full size table

We used Student’s t-test to assess differences in age, WOMAC scores, JSW, Tönnis angle, and dGEMRIC indexes measured preoperatively for patients whose hips were preserved at the time of last followup and those who failed. The nonparametric Mann-Whitney U test was used to compare LCE and ACE angles since these radiographic measurements did not follow a normal distribution and medians and interquartile ranges are reported. We used the Pearson chi-square test to compare ordinal data between the two groups including joint congruency and Tönnis grade and Fisher’s exact test to compare sex and left versus right side. Correlation between the coronal and sagittal dGEMRIC index values was measured using the Pearson product-moment correlation coefficient (r) with a fitted linear regression line and 95% CI for individual hips. In preparation for a multivariate analysis, a univariate analysis was performed to identify the relevant variables. Univariate analysis demonstrated preoperative pain, JSW, Tönnis grade, and coronal and sagittal dGEMRIC indexes were different between failed and preserved hips (Table 1). We used multivariable logistic regression analysis with backward stepwise selection to assess the influence of WOMAC pain, JSW, Tönnis grade, and different permutations of dGEMRIC indexes to control for confounding, with the goal of identifying independent predictors of failure after PAO using the likelihood ratio test (LRT) to assess level of significance. Receiver operating characteristic (ROC) curve analysis was utilized to evaluate the quality of outcome prediction using the dGEMRIC index with area under the curve (AUC) as a measure of accuracy and to determine an optimal cutoff value. We performed data analysis using the SPSS® package (Version 19.0; SPSS Inc, Chicago, IL, USA). Two-tailed values of p < 0.05 were considered significant.

Results

The mean dGEMRIC indexes of coronal images correlated (r = 0.742, r2 = 0.551, p < 0.001) with those of the sagittal images (Fig. 1). The mean dGEMRIC indexes of the average coronal (p < 0.001) and sagittal (p < 0.001) planes and their individual slices were lower in the prematurely failed group (Table 2). The sagittal slices were divided further into anterior and posterior weightbearing regions. The mean dGEMRIC indexes for both the anterior and posterior weightbearing regions also differed between the two groups, but the difference was greater in the posterior weightbearing region (Table 2). In particular, the posterior weightbearing region in the failed group had one of the lowest dGEMRIC indexes.

Fig. 1A–C
figure 1

Sample (A) coronal and (B) sagittal dGEMRIC images are shown. (C) A graph shows the correlation of the average sagittal dGEMRIC index with the average coronal dGEMRIC index.

Full size image
Table 2 Comparison of regional dGEMRIC indexes from coronal and sagittal averages and individual slices
Full size table

WOMAC pain (LRT = 5.28, p = 0.022) and sagittal anterior dGEMRIC index (LRT = 20.45, p < 0.0001) independently predicted failure after PAO. The other covariates tested provided no additional value in predicting failure (all p > 0.20). The predictive value of dGEMRIC can be best depicted by a graph showing the increased probability of failure after PAO with shorter sagittal anterior weightbearing dGEMRIC times (Fig. 2). For example, the theoretical probability of failure for a hip with a dGEMRIC of 250 milliseconds is 96% (95% CI, 63%–99%). ROC analysis revealed the sagittal anterior weightbearing dGEMRIC provided excellent discrimination between hips that failed after PAO and those having a satisfactory outcome with an AUC of 0.977 (95% CI, 0.938–1.000) (Fig. 3). The optimal cutoff value based on ROC analysis was 370 milliseconds, which was associated with a sensitivity of 100% (all six failures had values ≤ 370 milliseconds) and a specificity of 95% (35 of the 37 satisfactory outcomes were in hips with an average value > 370 milliseconds).

Fig. 2
figure 2

A graph shows the calculated theoretical probability of failure after PAO using the average sagittal anterior weightbearing dGEMRIC index. The probability of a failed PAO decreases sharply with increasing dGEMRIC time as illustrated by the curve. The dashed curves represent the 95% CI as an indication of precision of the estimated probability.

Full size image
Fig. 3
figure 3

An ROC curve shows excellent discrimination of the average sagittal anterior weightbearing dGEMRIC in predicting failures (sensitivity) and nonfailures (specificity) after PAO. The optimal cutoff value is 370 milliseconds.

Full size image

Discussion

At present, several different compositional MRIs are available, of which some have been validated in the clinical setting to detect early OA. dGEMRIC has been most frequently used in clinical studies, and we previously reported it predicts the need for early joint arthroplasty after Bernese PAOs. In this study, we presumed regional variations in mechanical stress in a dysplastic hip would lead to degenerative changes that are nonuniform. We believed identifying areas with the earliest cartilage damage could be used as an improved predictor over the dGEMRIC index on coronal slices and plain radiographic measure of OA.

Our study is associated with a number of limitations. First, our sample size was limited and especially the number of failures was small. In our prior study, the number of failures was larger; however, many did not have sagittal dGEMRIC scans and therefore were not used. This may explain why joint subluxation was no longer a predictor of premature joint failure. Similarly, the small number of failures will limit the validity of a multivariable analysis. Therefore, we would interpret our results as preliminary and encourage further work. Second, the length of followup in this study was short. We believe dGEMRIC is certainly a valuable predictor of short-term premature joint failure after PAO; however, further study is necessary to see whether dGEMRIC is a useful predictor of long-term joint preservation. Third, dGEMRIC is a new MRI technique unfamiliar to most imaging centers. To replicate our findings, familiarity with this imaging technique and the potential technical limitations is recommended.

Regional variations in biochemical content of cartilage have been described in other papers. In FAI, global arthritic changes were noted in pincer-type hips and more anterior and central in cam-type hips [38]. In patients with FAI, with or without symptoms of impingement, higher alpha angle correlated with reduced dGEMRIC index on the anterosuperior aspect of the hip and more globally in patients with FAI with signs of impingement [58]. Additionally, we know, in dysplastic hips, often the articular cartilage damage first occurs in the anterior portion of the acetabulum [66]. Our analysis of the individual slices of coronal and sagittal dGEMRIC indexes demonstrated the amount of degeneration in these individual areas correlated with overall lower dGEMRIC indexes. However, the damage was most noticeable in the anterior portion of the sagittal cuts, consistent with previously reported findings [53].

Given there is more extensive damage in the anterior portion of the acetabulum, we sought to know whether the dGEMRIC index in this region would best predict early joint failure after PAO. It was interesting, despite the facts that the posterior part of the joint had a larger difference in dGEMRIC index between the failed and successful group and that coronal and sagittal dGEMRIC values correlated to each other, multiple logistic regression models comparing these indexes revealed the sagittal anterior weightbearing region dGEMRIC index best predicted early joint failure. One explanation for this result could be the anteversion of the femoral head causing greater mechanical load in the anterior region of a dysplastic hip. Murphy et al. [48] reported variable amounts of acetabular deficiency in patients with dysplasia, with some having more global dysplasia and others with regional dysplasia. Additionally, the mechanical load and stressors of each individual’s hip are unique depending on the amount of version of the acetabulum and femoral head and the congruity. Perhaps even after acetabular redirection, the anterior part of the joint is still under higher mechanical load and hence excessive damage in this part of joint will lead to continued degeneration.

In summary, biochemical imaging can show early damage in the hip consistent with what we understand from the pathomechanics of hip deformity. In surgical correction of acetabular dysplasia, it appears dGEMRIC may have a role in selecting the right patient for treatment. Further work is needed to see whether other biochemical imaging techniques will have the same clinical value, especially in conditions such as FAI. The promise of a biochemical imaging technique is its potential to demonstrate, early after treatment, changes in cartilage health that may indicate a procedure may help preserve a joint for decades to come. Ultimately, that is the goal of joint preservation surgeries.