Radiographic-Based Measurement of Tibiofemoral Joint Space Width and Magnetic Resonance Imaging Derived Articular Cartilage Thickness are Not Related in Subjects at Risk for Post Traumatic Arthritis of the Knee

Abstract

Joint space width (JSW), measured as the distance between the femoral and tibial subchondral bone margins on two-dimensional weight-bearing radiographs, is the initial imaging modality used in clinical settings to diagnose and evaluate the progression of osteoarthritis (OA). While, JSW is the only structural outcome approved by the FDA for studying the treatment of this disease in phase III clinical trials, recent reports suggest that magnetic resonance imaging (MRI)-based measurements of OA changes are superior due to increased sensitivity and specificity to the structural changes associated with progression of this disease. In the current study, we examined the relationship between radiographic JSW and MRI-derived articular cartilage thickness in subjects 4 years post anterior cruciate ligament reconstruction (ACLR) who were at increased risk for the onset and early progression of post-traumatic OA, and in uninjured subjects with normal knees (Control). In both ACLR and Control groups, there were large measurement biases, wide limits of agreement, and poor correlation between the two measurement techniques. Clinical significance: The finding from this study suggest that the two methods of examining changes associated with the onset and early progression of PTOA either characterize different structures about the knee and should not be used interchangeably, or two-dimensional JSW measurements are not sensitive to small changes in articular cartilage thickness.

Joint space width (JSW) measurements obtained from plain film radiography have become a common research assessment for characterizing the progression of post-traumatic osteoarthritis (PTOA) because of its affordability, availability, capacity to assess the joint during weightbearing conditions and prior experience with using it to study primary osteoarthritis (OA).¹ By measuring the projected distance between the subchondral bone margins of the femur and the tibia, JSW indirectly quantifies the articular and meniscal cartilages as a composite structure; however, the sensitivity of this measure to quantify changes in these structures is considered poor (7–46%).²

Focus has turned toward the use of magnetic resonance imaging (MRI) to characterize the structures that undergo change associated with the onset and progression of PTOA. While both radiographic and MRI measurement approaches have been lauded as superior for evaluating the progression of OA and PTOA,^3,4 MRI allows for direct three-dimensional measurement of cartilage thickness, complete visualization of the cartilage and menisci, and accurate mapping of femoral, tibial, and patella surfaces in all compartments of the knee. Thus, it is believed that MRI may provide more specific, sensitive, and reliable information on the structural changes that occur about the entire knee as OA and PTOA progress over time.

It is unclear whether two-dimensional radiographic-based measurements of JSW and three-dimensional MRI-based measurements of articular cartilage thickness are indeed related in subjects at risk for developing PTOA. This served as the motivation for the current study which was to compare radiographic JSW and MRI articular cartilage thickness measurements in a cohort of males and females who underwent anterior cruciate ligament reconstruction (ACLR) and returned to pre-injury activity as well as a group of subjects with normal knees. The aim of this study was to determine if radiographic JSW and MRI-derived articular cartilage thickness measurements are related in subjects at increased risk for PTOA and subjects with normal knees.

METHODS

Patient Population and Study Design

This study was approved by the University of Vermont’s Institutional Review Board and was based on secondary analysis of image-based data obtained from a longitudinal cohort study that evaluated the relationship between collagen markers and structural changes in the knee following ACL reconstruction and return to activity.⁵ Participants were recruited into the study from a community-based orthopedic clinic at the time of their first ACL tear and included based upon their age (14–55 years), BMI (18.5–30), and Tegner activity level (greater than 5). They were excluded if they had previous injuries or surgeries to either knee, radiographic evidence of OA, fracture, abnormal alignment of the lower limb, articular cartilage lesions of ICRS grade 3A or greater, or a meniscus tear that required resection of more than one-third of the inner portion of the structure. ACL reconstructions were performed with bone-patellar tendon-bone autografts by one of two sports medicine fellowship trained orthopedic surgeons and all subjects participated in the same rehabilitation program.⁶ Concurrently, control subjects were recruited from the surrounding community and included if they did not report knee dysfunction or pain, had no medical history of significant trauma to any diarthrodial joint, normal clinical knee exam, and normal MRI findings at baseline and the 4 year follow-up.⁵ The ACL injured subjects and control subjects were similar with regard to body mass index, age, activity level, and race.

Radiographic measurement of Tibiofemoral Joint Space Width (JSW)

Bilateral posterior-anterior weight-bearing radiographs of the knee were obtained with a fluoroscopy assisted, semi-flexed, technique that has been shown to provide accurate and reliable measurements of JSW.⁴ Radiographs were digitized using a Umax UTA-2100XL Scanner (Umax Technologies, Inc., Dallas, TX) and JSW was evaluated by the same examiner for medial and lateral compartments using a previously described midpoint technique.⁷ For each compartment of the knee, the midpoint technique involved constructing two parasagittal reference lines, one directed through the apex of the tibial spine and the other through the outermost periphery of the compartment, and then constructing a midpoint line that was equidistant between and parallel with the reference lines. The intersection of the midpoint line with the subchondral bone of the tibia and femur were the location of the measurement points.

MRI Measurement of Articular Cartilage Thickness

Cartilage thickness profiles were generated from 3T MRIs that were obtained from the same scanner and acquired by the same technician (Philips Achieva, Philips Healthcare, Best, the Netherlands).⁶ The sagittal plane, T1 weighted, Fast-Field Echo scans (within-plane resolution of 0.3mm by 0.3mm and a slice thickness of 1.2mm) were acquired with the patients supine using an 8-channel SENSE coil. Articular cartilage thickness maps were established for the tibia and femur. Cartilage thickness measurements were calculated from manually segmented articular cartilage surfaces and the subchondral bone-cartilage interface surfaces using a previously described technique that has been shown to provide accurate and reliable measurements.^8–10 The MRI-based thickness measurements were made at same location as the radiographic JSW measurements and this was accomplished by acquiring data at the medial and lateral local minima of the tibial articular cartilage surfaces and at the corresponding locations on the femur. In each compartment, articular cartilage thickness was calculated as the sum of the tibial and femoral cartilage thicknesses.

Statistical Analyses

First the radiographic JSW measurements were plotted against the MRI-derived articular cartilage thickness measurements. This was done for; each compartment of the knee; cases and controls; and for the females and males. We then constructed Bland-Altman plots and determined the difference between the two measurements (bias) with the corresponding lower and upper limits of agreement. In addition, linear regression of radiographic JSW on MRI cartilage thickness was performed and the Pearson product-moment correlation coefficient (r), the coefficient of determination (R²), and the slope and intercept of the relationship used to characterize the relationship and quantify its strength. For the ACLR and control groups, this was done separately for each sex and for each compartment of the knee.

Additional analysis was conducted to replicate an experimental design that would use subjects as their own control, minimizing between subject variability associated with the measurements. This involved calculating the side-to-side differences of the measurements (i.e., subtracting the value measured in the uninjured knee from the value measured in the ACLR knee). The side-to-side JSW differences data were plotted against the side-to-side MRI articular cartilage thickness difference data and the Pearson correlation analysis was performed. For the ACLR and control groups, this was done separately for each sex and each compartment of the knee.

RESULTS

Comparison of Radiographic JSW and MRI Articular Cartilage Thickness Values

Bilateral radiographs and MRIs acquired at the 4-year follow-up interval from 23 subjects that underwent ACLR (10 male and 13 female; Table (1) and 10 control subjects (5 males that had an average age of 35 ± 4.5 and BMI of 28 ± 11.9, and 5 females that had an average age of 37 ± 8.3 and BMI of 22 ± 0.4) were analyzed. Subjects in the ACLR group either did not have radiographic evidence of PTOA (KL grade 0), or very minor changes (KL grade 1 or (2) (Table 1). The controls did not have radiographic evidence of PTOA (KL grade 0). Subjects in the ACLR and control groups participated in sports and activities without pain and symptoms (Table 1). All subjects in the ACLR group returned to their pre-injury activates and sport at the pre-injury activity level, and were not limited by their injury and reconstruction (Table 1).

Table 1.

Anterior Cruciate Ligament Reconstruction (ACLR) Cases Anthropometric Data, Knee Osteoarthritis Outcome Scores (KOOS), and Kellgren-Lawrence (KL) Grades at Time of Four Year follow-up

			KOOS
			(Mean±Standard Deviation)					KL Grade Frequency
	Age (years)	BMI (kg/m2)	Symptoms	Pain	Activities of Daily Living	Sports	Quality of Life	0	1	2	3
Females	30±12.8	23±2.2	87±13.9	93±12.1	97±5.5	90±11.5	87±12.6	8	5	0	0
males	34±11.2	25±2.5	94±7.4	96±5.3	100±0.4	96±6.1	89±13.1	7	2	1	0

The Bland–Altman analysis estimated the bias and limits of agreement associated with the measurement techniques for the: ACLR and control group; medial and lateral compartments of the knee; and males and females (Figs. 1–4). The bias was almost twofold larger in the medial compartment of the knee in the ACLR group compared to the same compartment of the controls while the limits of agreement were similar. In contrast, for the lateral compartment, the bias was larger in the control group in comparison to the ACLR group while the span in the limits of agreement were twofold larger for the ACLR group. When considering the control group, the bias associated with the measurement techniques differed between the medial and lateral compartments of the knee while the spans in the limits of agreement were similar. (Table 2, Figures 1 and 2). For the medial compartment, MRI-based measures of cartilage thickness were greater than the corresponding radiographic measures of JSW. In contrast, for the lateral compartment, MRI-based measures of cartilage thickness were less than the radiographic measures of JSW. In addition, for the control group the bias between the measurement techniques was similar in magnitude and direction between the males and females in both the medial (0.63 vs. 0.36) and lateral compartments (−0.67 vs. −0.62). In contrast, for the ACLR group, the bias between the measurement techniques differed between the males and females (1.25 vs. 0.51 and 0.27 vs. − 0.32) for the medial and lateral compartments, respectively (Table 2, Figures 3 and 4). In addition, the spans in the limits of agreement for both compartments were larger for males than females.

Figure 1.

Bland Altman analysis are presented for the lateral compartment of the control subjects knee. Included is the mean difference between the measurement techniques (bias), the corresponding upper and lower limits of agreement, and the data acquired from the male and female subjects.

Figure 4.

Bland Altman analysis of data acquired from the lateral compartment of subjects in the ACLR group. Included is the mean difference between the measurement techniques (bias), the corresponding upper and lower limits of agreement. Data are presented for the male and female subjects.

Table 2.

Radiographic JSW and MRI Articular Cartilage Data From the ACLR and Control Groups are Presented as Average Values With 95% Confidence Intervals (CI) for Each Compartment, and Each Sex

	MRI Cartilage Thickness AVE ± 95 CI	Radiographic JSW AVE ± 95% CI	Diff	Bland–Altman Lower Limit of Agreement	Bland–Altman Upper Limit of Agreement	R2	Slope	Intercept
	(mm)	(mm)	(mm)	(mm)	(mm)			(mm)
ACLR - Cases
Female
Medial	3.96±0.30	3.45±0.25	0.51	−0.44	1.46	0.61	0.66	0.82
Lateral	4.68 ± 0.42	5.00 ± 0.27	−0.32	−1.89	1.26	0.45	0.43	3.0
Male
Medial	4.89 ± 0.32	3.64 ± 0.36	1.25	−0.51	3.01	0.12	0.83	1.79
Lateral	5.82 ± 0.49	5.55 ± 0.43	0.27	−2.64	3.17	0.1	0.1	5.55
Controls
Female
Medial	3.68 ± 0.45	3.32 ± 0.31	0.36	−0.51	1.22	0.64	0.55	1.31
Lateral	4.49 ± 0.49	5.11 ± 0.39	−0.62	−1.77	0.53	0.47	0.54	2.68
Male
Medial	4.62 ± 0.40	3.99 ± 0.19	0.63	−0.68	1.95	0.03	0.08	3.62
Lateral	5.16 ± 0.30	5.83 ± 0.30	−0.67	−1.83	0.49	0.07	0.27	4.46

The findings from the Bland–Altman analysis are presented as the average difference (bias) between the radiographic JSW and MRI cartilage thickness with the corresponding lower and upper limits of agreement for each sex and each compartment. In addition, the coefficient of determination, the (R²), and linear regression coefficients between radiographic JSW and MRI-based measurement of articular cartilage thickness in the ACLR and control groups are presented.

Figure 2.

Bland–Altman analysis of data acquired from the medial compartment of the control subjects. Included is the mean difference between the measurement techniques (bias), the corresponding upper and lower limits of agreement. Data are presented for the male and female subjects separately.

Figure 3.

Bland Altman analysis of data acquired from the lateral compartment of subjects in the ACLR group. Included is the mean difference between the measurement techniques (bias), the corresponding upper and lower limits of agreement. Data are presented for the male and female subjects separately.

There were moderate correlations between radiographic JSW and MRI cartilage thickness measurements for females in the ACLR group for the medial (R² = 0.61) and lateral compartments (R² = 0.45). In contrast, there were poor correlations between the measurement techniques for both compartments of the males in the ACLR group (R² = 0.12 and 0.1 for the medial and lateral compartments, respectively).

Analysis of the data acquired from the subjects in the control group revealed similar findings. Comparison of the measurements resulted in moderate correlations for the female controls in the medial (R² = 0.64) and in the lateral compartments (R² = 0.47), while there were poor correlations among the male controls for both compartments (R² = 0.03 and R² = 0.07 for the medial and lateral compartments, correspondingly).

Comparison of Radiographic JSW and MRI Measurements of Side-to-Side Differences in Articular Cartilage Thickness

With regard to the side-to-side differences analysis, comparison of the measurement techniques for the female ACLR subjects showed poor correlations in the medial (R² = 0.11) and lateral compartments (R² = 0.15). Similarly, there were moderate to poor correlations for the male ACLR subjects in the medial (R² = 0.30) and lateral compartments (R² = 0.17). For female control subjects, comparison of the side-to-side differences in the measurements resulted in poor correlations for the medial compartment (R² = 0.07) while there was an improvement in the correlation for the lateral compartment (R² = 0.56). For male control subjects, there were poor correlations for the medial (R² = 0.06) and lateral compartments (R² = 0.02).

DISCUSSION

This investigation revealed large bias, wide limits of agreement, and poor to moderate correlations between radiographic JSW and MRI measurements of cartilage in both the ACLR and control groups. This suggests that each technique may evaluate different structures and that they should not be used interchangeably in studies of the onset and early progression of PTOA. This finding confirms the prior study of females with advanced stages of OA reported by Marsh et al.¹¹ that revealed JSW and MRI evaluate different anatomic structures in the knee joint. They highlighted that the MRI articular cartilage measurement technique is typically performed with the knee unweighted and does not include the menisci, while the radiographic JSW technique is performed during weight bearing and includes both the articular cartilage and meniscal contributions to the JSW. They also found that when both techniques were performed during weight-bearing conditions, there was not a statistically significant improvement in correlation between the measurement techniques. This led them to conclude that the weight-bearing condition of the knee during radiograph acquisition does not alone account for an imperfect correlation between JSW and MRI articular cartilage thickness.¹¹ With this observation in mind, it is important for us to point out that the JSW measurements that were made in the current study were obtained from weight-bearing radiographs, while the MRI based measured were made while the knee was unloaded.

Our study also revealed that the correlations between JSW and MRI-based cartilage thickness measurements were not consistent between males and females. For the female ACLR and control groups, the comparisons between the measurement techniques produced correlations that demonstrated moderate strength with 61–64%, and 45–47% of the variability between the measurement techniques explained for the medial and lateral compartments, respectively. The strength of these correlations are similar to study of female subjects with OA reported by Marsh et. al.¹¹ In contrast, these relationships were consistently poor for the male ACLR and control groups, with no more than12 percent of the variability explained for the medial and lateral compartments. This sex-specific response may be explained, at least in part, by the fact that a larger portion of the males in our study underwent partial meniscectomy at the time of ACL reconstruction. Of the 10 male cases, 6 had 15–30% of the meniscus removed (4 in the lateral compartment) while for 13 female cases, 3 underwent partial meniscectomy, with only 5 to10% loss of the meniscus (2 in the lateral compartment). The subjects that underwent partial meniscectomy, only had a small proportion of the meniscus removed; however, it may be that the location and extent of meniscal disruption, and the amount of tissue resected has a direct effect on the JSW measurement while not effecting the MRI-based measurement of cartilage thickness. This may have contributed to the large bias and highly variable correlations that were observed in the measurements that were made on the males.¹¹ It is important for us to point out that we did not include subjects that underwent complete meniscectomy; instead, we included subjects that had very minor injuries to the meniscus that underwent partial resection (one subject had less than one third of the inner portion of the meniscus removed and a majority had less than 15% of the meniscus removed). As well, the subjects had very minor articular cartilage lesions. This approach was used to allow translation of the findings from the study to the most common injuries associated with severe knee ligament injury (i.e., those with ACL and meniscus injuries) and consequently the most common group that is at increased risk of developing PTOA following severe knee ligament trauma.

The measurement techniques used in this investigation were based on prior reports that focused on studying the progression of primary OA and PTOA. Our approach was to compare these measurement techniques in subjects at risk for developing PTOA during the early stages of disease progression as very little is known about the biological and structural processes associated with the onset of this disease. At the 4 year follow-up, one-third of the subjects in the ACLR group underwent a significant change in JSW;⁴ however, they maintained the same function and activity level compared to their pre-injury values and consequently it is reasonable to assume that the subjects used their knees in a similar manner over time.⁴ We did observe an increase in anterior-posterior laxity in all subjects in the ACLR group; however, none of the subjects reported giving-way episodes at the knee, pivot shift episodes, or limitations during participation in sport and activities of daily living. From this perspective the findings from our study pertain to a relatively homogeneous group of subjects with normal alignment of their lower extremities that suffered an ACL injury with and without minor concomitant injury to the meniscus and did not have significant articular cartilage lesions, but are at increased risk for development of PTOA. It is unclear if the findings from the current study apply to subjects with abnormal alignment of their lower extremity that suffer ACL injury, undergo reconstruction and return to their pre-injury activities, and this should be the focus of future investigations. Also, the MRI-based measurement of femoral and tibial articular cartilage thickness maps were acquired with an approach that used manual segmentation and this was exceedingly time consuming; consequently, our first step was to conduct a correlative study that included a relatively small number of subjects. This approach did not provide adequate power for testing whether agreement between the two measurement techniques differed significantly between males and females, or between the control and ACLR groups.

An important strength of our study was the use of a prospective study design that included subjects that suffered ACL trauma and control subjects that were matched based on activity level, age, sex, and body mass index. All study participants were active (Tegner score of 5 or greater). In addition, we excluded subjects with severe concomitant injuries (i.e., subjects did not have concomitant injuries to the other primary ligaments of the knee, they had normal articular cartilage, and only minor injuries to the meniscus), previous injury to any diarthrodial joint, prior knee injections, or varus/valgus mal-alignment. Further, the radiographs and MRI data were acquired within an hour of each other during the same visit. The medial and lateral compartments were considered in this study because both compartments are at increased risk of developing PTOA following ACL injury and reconstruction.^6,7,8

Prior to conducting this study the accuracy and reliability of the measurement techniques were established. The accuracy associated with the radiographic JSW measure was 0.13mm,⁷ while the accuracy associated with the MRI-based measure was 0.15mm.⁸ Similarly, the reliability of the JSW and MRI-based measures between repeated measurement sessions were considered to be very good (0.11mm and ICCs of 0.88 greater).^7,8 These accuracy and reliability values are far less than the bias and limits of agreement that were found in the current study and consequently our findings should be considered biomechanically significant.

At the current point in time very little is known about the biomechanical and biological processes associated with structural changes to the knee during the early stage of PTOA progression and this has served as the underlying motivation for using different image-based techniques to gain insight into these processes. In this study two popular image based techniques that have been applied to study PTOA, radiographic JSW, and MRI articular cartilage thickness measures, were compared to determine if they are related in patients that suffered ACL rupture, underwent reconstruction, and were at increased risk of developing PTOA. Our study participants were not symptomatic and did not have radiographic evidence of PTOA (Table 1). We found large measurement bias and wide limits of agreement between the two measurement techniques for both groups of subjects. In addition, the correlation between the two methods of quantifying change within the knee joint ranged between poor to moderate and poor. These findings are important to consider when using image-based techniques to evaluate the initial change of PTOA in subjects that have suffered an ACL injury and have not experienced symptoms of the disease, as they suggest each measurement technique may provide different insight into the onset and early progression of this disease and measure change to different structures in the knee, or that two-dimensional JSW measurements are not sensitive to small changes in articular cartilage thickness.