Subchondral Trabecular Bone Integrity Changes Following ACL Injury and Reconstruction: A Cohort Study with a Nested, Matched Case-Control Analysis

Abstract

OBJECTIVE

There is limited information regarding changes in bone architecture following ACL injury. The objective of this study was to evaluate differences in tibial fractal signature in the medial and lateral compartments following ACL injury and describe how these values change following ACL-reconstruction and return to activity.

DESIGN

This was a prospective cohort study with a nested case-control analysis. ACL-injured subjects and matched controls were evaluated at pre-surgical baseline and post ACL reconstruction follow-up at a mean of 46 months. Serial Fractal Dimensions (FD) of tibial subchondral bone architecture were calculated in medial and lateral regions of interest in the horizontal and vertical dimensions.

RESULTS

In the medial and lateral compartments, there were significant differences in the vertical FD signature for ACL-injured subjects at final follow-up, when compared to the contralateral healthy tibia (medial p<0.0001; lateral p<0.0001) and the control group (medial p=0.01; lateral p<0.0001). Similarly, in the lateral compartment, there were significant differences in the horizontal FD profile for ACL-injured subjects at final follow-up, when compared to the contralateral healthy tibia (p=0.003) and the controls (p<0.0001). There were no significant side-to-side differences in FDs among healthy control subjects in the medial or lateral compartments at baseline or final follow-up.

CONCLUSION

At 46-month follow-up, Fractal Dimension profiles are significantly different, and show an overall lower FD signature, for ACL-injured knees when compared to the contralateral healthy knee and uninjured controls. Additionally, this study provided the first side-to-side symmetry data of medial and lateral FD values in healthy controls.

Introduction

Identifying the earliest stages of onset of PTOA provides the greatest opportunity for prevention and treatment of this disease, prior to substantial articular cartilage and bone changes. By definition, patients suffering from PTOA have experienced acute trauma and can be observed early in the course of their disease. It is for this reason that the Osteoarthritis Research Society International (OARSI) FDA Osteoarthritis Biomarkers Working Group has recommended targeting post-traumatic osteoarthritis patients in future studies.(1)

It has been well established that significant knee trauma such as rupture of the anterior cruciate ligament (ACL) is associated with the development of post-traumatic osteoarthritis (PTOA),(2) with radiographic changes presenting in 60–90% of patients within 10–15 years of injury.(3–5) Despite attempts at restoring normal joint biomechanics through surgical reconstruction of the ligament, PTOA still develops in most ACL-reconstructed individuals.(6) The mechanism of early-onset PTOA in the ACL-injured knee remains unknown, and its pathogenesis is not well understood.(7)

Changes in subchondral bone are an integral part of OA pathogenesis.(8–11) Subchondral bone progresses from a resorptive pattern early in the disease process to a sclerotic pattern in later phases, as demonstrated by imaging(12) and with validated biochemical markers of bone resorption.(13) Furthermore, longitudinal studies of knee OA have shown an interdependent relationship between subchondral bone changes and articular cartilage loss.(14, 15) The cause and effect relationship between subchondral bone changes and articular cartilage loss has yet to be determined in humans, but multiple animal models have demonstrated that subchondral bone changes precede articular cartilage changes.(16, 17)

Fractal signature analysis (FSA) is an alternative methodology that utilizes plain radiographs to measure the trabecular structure of bone. First described by Lynch and colleagues in the early 1990s,(18, 19) FSA quantifies the microscopic boney anatomy into Fractal Dimensions (FD) using 2-D radiographs at a variety of scales within the range of the trabecular anatomy. This technique has more recently been utilized in the context of knee OA evaluation and potentially for predicative modeling (20–24). The FD represents the number and spacing of trabeculae in the horizontal and vertical planes of the radiograph, which have been shown to represent the pore size of the bone marrow.(25)

According to Messent et al(26), a decrease in the horizontal and vertical fractal signature represents a decrease in image complexity, which can be caused by two distinct processes: either thickening of individual trabeculae or a total loss of trabeculae. In the proximal tibia in the setting of OA, the former represents subchondral sclerosis and the latter represents subarticular osteopenia as a result of stress shielding. This group has hypothesized that subchondral sclerosis of the tibia plateau may be associated with redistribution of the contact stresses that are transmitted across the tibiofemoral joint, reducing load transmission to the deeper trabecular bone, which may result in local osteopenia.(27) It is not clear whether the initial insult provided at the time of injury, surgery, or rehabilitation is responsible for this potential change, or if the resultant changes in joint biomechanics and reduced loading are implicated from a mechanistic perspective. This study was not designed to evaluate these potential causative factors, which should be the subject of future investigation.

Currently, there is limited data regarding changes in the subchondral trabeculae of the proximal tibia following severe joint trauma such as anterior cruciate ligament disruption. Furthermore, it is not known if “baseline” FD measurements of the ACL-injured knee that is observed soon after ACL trauma are in fact within normal limits and are appropriate to use for the assessment of FD change over time. Additionally, it is also not known if FD measurements obtained in healthy knees are the same bilaterally (right vs. left), or if they change over time.

The purpose of this investigation was to evaluate differences in medial and lateral subarticular tibial FD values following ACL injury, surgical reconstruction, and rehabilitation and describe how these values change over time (two time points, pre-surgical baseline compared to 4-year post-surgical follow-up) compared to the contralateral knee and to healthy, non-injured controls. Three hypothesis were tested and these included: 1) There are no significant within subject side-to-side (injured vs non-injured) differences in FD for ACL-injured subjects soon after injury at pre-surgical baseline; 2) Differences in the fractal signature do not change significantly over time in ACL-R subjects compared to the contralateral knee and to healthy, non-injured controls; and 3) There are no statistically significant side-to-side differences in fractal signature within control subjects with normal knees.

Methods

Study Design and subject population

This study was secondary analysis of data that was obtained as part of a prospective cohort study designed to study the onset of PTOA and involved 32 first-time, acute ACL-injured subjects (18 women), and 32 healthy controls (18 women) matched by sex, age, BMI, and pre-injury activity level (Table 1).(28) These patients were selected from the larger cohort and details of the surgical procedure have been described.(29, 30) Entry criteria for injured subjects included: age between 14–55yrs; BMI between 18.5–30, Tegner activity score greater than or equal to 5; no previous joint surgeries (to any joint) or knee injections; no relevant knee pathologies other than the index injury; no abnormal clinical laxity of any ligament other than the ACL; no evidence of arthritis on baseline weight bearing posterior-anterior metatarsal phalangeal view radiographs, no obvious varus or valgus malalignment (2000 IKDC Knee Examination Criteria); less than 2/3 medial or lateral meniscectomy; and grade 3A or less articular cartilage lesions of tibiofemoral or patellofemoral joints as proposed by the International Cartilage Repair Society (ICRS), documented arthroscopically by a fellowship-trained surgeon.

Table 1.

Patient Demographics

	Cases (n =32; mean (SD))	Controls (n=32; mean (SD))
Age (yrs)	29.4 (+/−12.6)	26.7 (+/−6.6)
Sex	18 F (56%)	18 F (56%)
BMI	24.8 (+/− 3.8)	23.4 (+/− 3.3)
Tegner	3 (+/− 1.5)	6.2 (+/− 1.3)

Yrs, Years; F, female; BMI, Body Mass Index; SD Standard Deviation

Entry criteria for control subjects were similar to injured participants with the following exceptions: no subjective report of knee pain or dysfunction as determined by the Knee Injury and Osteoarthritis Outcome Score (KOOS)(31) and International Knee Documentation Committee (IKDC) subjective knee evaluation(32); no history of significant joint trauma (defined as that requiring physician referral and/or more than three days modified activities of daily living); no abnormal findings with clinical knee examination (IKDC); and no abnormal findings on baseline or follow-up of either knee. Our Institutional Review Board approved the study. All subjects provided informed consent to participate prior to enrollment.

Surgical procedure and rehabilitation

The ACL reconstructions were performed by two fellowship trained orthopaedic surgeons. ACL-injured subjects underwent arthroscopic reconstruction within six months of their injury (mean time between index injury and surgery date was 66.6 days (SD = 29.6), range: 18–155 days) and all subjects participated in the same standardized post-surgical rehabilitation program.(29) Graft choice was at the discretion of the surgeon. The tension applied to the graft at the time of fixation was standardized with a KT-1000S arthrometer (MedMetric Corp, California) using a technique that has been described.(29) Injured subjects were assessed at baseline (within 3 weeks prior to surgery), and at a mean (SD) follow-up time of 46.1 (+/−8.6) months. Matched control subjects also underwent baseline and follow-up evaluations at a mean of 33 months (SD = +/−6.6; Range = 13.3 – 44.73)) months. Control subjects underwent baseline and follow-up 3T MRI to confirm that no pathology was present and ensure no injuries were incurred during the study period. ACL-group subjects also underwent bilateral 3T MRI at follow-up.

Radiographic procedures

Bilateral posterior-anterior (PA) view knee radiographs were obtained on all subjects during each visit using a fluoroscopy-assisted semi-flexed, metatarsal phalangeal (MTP) view technique following the approach that has been previously described.(28, 33, 34)

Image Analysis

Radiographs were digitized with a Umax UTA-2100XL transparency scanner (Umax Technologies, Inc, Dallas, Texas) at resolution of 400×400 pixels/in and a corresponding pixel size of 0.06mm as has been utilized in previous work (35). Medial and lateral compartment FSA was performed using commercially available software (KneeAnalyzer, Optasia Medical, Manchester, UK). The medial tibial joint profile was defined using a 6-point manual initialization (medial and lateral femur, medial and lateral tibial, and tibial spines). Based on the medial tibial joint profile, a region of interest (ROI) in the medial subchondral bone was defined for FSA. The ROI had a height of 6mm and included three fourths of the width of the medial tibial compartment, with its medial most border aligned with the medial tibial spine and was standardized by the software, based on prior methodology(20) and replicated by others.(21) The radiograph was flipped horizontally and these measurements were re-calculated, defining a subchondral area of interest in the lateral compartment of the tibia. Intra-class correlation coefficients (ICC) were calculated for intra-observer reliability of vertical and horizontal FD measurements. This included the complete re-evaluation of ROI and re-calculation of FSA with a minimum of 1 week between analyses for the same observer. Calculated intra-observer ICCs showed strong correlations, with values greater than 0.8.

Statistical analysis

For fractal signature analyses, raw horizontal and vertical FD data were used, in order to model the trends of FD data in relation to the radius, and to test for side-to-side differences in the polynomial fit of these trends (i.e. differences in fractal dimension between sides), based on prior work by Kraus et al.(21, 36) A subject’s FD was modeled as a function of radius using a second order (quadratic) polynomial, due to the fact that initial regression analyses indicated that this was a good fit to the data. Hierarchical linear mixed models with repeated measures were used to fit the FD data, and compare polynomial fit between sides. In the mixed models, side, radius, and radius-squared (as these values varied between subjects) were fixed effects, while variation in the regression coefficients between individuals and deviations between the estimated and observed data points within the individuals were modeled as random effects. Interaction terms were included in the model as fixed effects to permit the regression coefficients to vary in a systematic manner between sides. Sex, age, BMI, time from injury to surgery (for baseline analyses), and time from baseline to follow-up (for final follow-up analyses) were used as covariates in the analysis models, in order to control for these factors. Comparisons were made at baseline, and at final follow-up, so that results can be compared at these two time points. In addition, time to surgery (for Baseline analyses) and time from baseline to final follow-up were used as covariates in the models in order to control for these effects. Interactions with time effects were not used in the poly fit mixed models, as the differences in fractal dimension profiles vs radius and radius-squared between groups were determined by comparing models with and without group interactions with these effects. Addition of time factor interactions would result in overly complex models that would be difficult to interpret.

Model parameters were estimated by maximum likelihood and the difference in the fit of models with and without the interaction terms was assessed by the likelihood ratio test to determine the significance of differences between sides in the polynomials used to characterize FD. For these analyses, left and right legs of control subjects were randomized to “injured” and “uninjured” sides, frequency matched to the same distribution of Injured/Uninjured vs. Left/Right as that of the ACL-R subject group. Similar mixed models were used, with fixed effect of group (ACL-R vs. Control), in order to test for significant differences in polynomial FD fit between groups, using “injured” leg data only. P-values will be presented to compare the polynomial fit mixed models, to determine if there are statistically significant differences between groups in the polynomial fit of the relationships between fractal dimension and radius/radius-squared. The p-value is derived by comparing the results of the likelihood ratio test between models with and without the group interactions present. Significance level alpha = 0.05 was used to determine statistical significance. Analyses were done using SAS 9.2 statistical software (SAS Institute, Inc. Cary, NC, USA).

Results

Of the 39 ACL-injured patients in the matched cohort, 7 patients were excluded from statistical analysis. Four did not have adequate follow up and 3 had poor image quality on follow up radiographs. Of the included radiographs, ACL-R hardware and surgical tunnels were not located within the medial or lateral FSA ROI. Descriptive statistics for medial and lateral FDs at each time point are presented in Table 2. P-values comparing significant differences in polynomial fit for each group (Figures 2 and 3) are presented in Table 3.

Table 2.

Results

Cases vs Controls: ACL-Injured vs Non-injured Subjects
Time point	Compartment	Cohort	FD Variable	Mean	Std. Dev.	Min.	Max.
Baseline	Medial	Cases	Vertical	2.489	0.215	2.080	3.072
		Controls	Vertical	2.529	0.207	2.104	3.125
		Cases	Horizontal	2.607	0.332	2.063	4.244
		Controls	Horizontal	2.576	0.254	2.112	4.194
	Lateral	Cases	Vertical	2.630	0.201	2.184	3.165
		Controls	Vertical	2.593	0.196	2.163	3.094
		Cases	Horizontal	2.765	0.320	2.122	4.206
		Controls	Horizontal	2.709	0.253	2.178	4.172
Final	Medial	Cases	Vertical	2.429	0.205	2.057	3.025
		Controls	Vertical	2.499	0.192	2.116	3.071
		Cases	Horizontal	2.592	0.311	2.101	4.296
		Controls	Horizontal	2.569	0.275	2.112	4.058
	Lateral	Cases	Vertical	2.582	0.185	2.165	3.034
		Controls	Vertical	2.583	0.206	2.148	3.068
		Cases	Horizontal	2.685	0.279	2.128	4.135
		Controls	Horizontal	2.692	0.270	2.135	4.310
Cases: ACL-injured Within Subject Side to Side Comparisons
Time point	Compartment	Side	FD Variable	Mean	Std. Dev.	Min.	Max.
Baseline	Medial	Injured	Vertical	2.489	0.215	2.080	3.072
		Uninjured	Vertical	2.493	0.220	2.081	3.100
		Injured	Horizontal	2.607	0.332	2.063	4.244
		Uninjured	Horizontal	2.599	0.337	2.066	4.285
	Lateral	Injured	Vertical	2.630	0.201	2.184	3.165
		Uninjured	Vertical	2.611	0.201	2.129	3.117
		Injured	Horizontal	2.765	0.320	2.122	4.206
		Uninjured	Horizontal	2.767	0.305	2.159	4.204
Final	Medial	Injured	Vertical	2.429	0.205	2.057	3.025
		Uninjured	Vertical	2.462	0.198	2.117	3.026
		Injured	Horizontal	2.592	0.311	2.101	4.296
		Uninjured	Horizontal	2.574	0.312	2.073	4.130
	Lateral	Injured	Vertical	2.582	0.185	2.165	3.034
		Uninjured	Vertical	2.602	0.195	2.190	3.082
		Injured	Horizontal	2.685	0.279	2.128	4.135
		Uninjured	Horizontal	2.709	0.301	2.154	4.228
Controls: Non-injured Subjects Side to Side Comparisons
Time point	Compartment	Side	FD Variable	Mean	Std. Dev.	Min.	Max.
Baseline	Medial	Injured	Vertical	2.529	0.207	2.104	3.125
		Uninjured	Vertical	2.525	0.209	2.146	3.103
		Injured	Horizontal	2.576	0.254	2.112	4.194
		Uninjured	Horizontal	2.573	0.257	2.162	3.988
	Lateral	Injured	Vertical	2.593	0.196	2.163	3.094
		Uninjured	Vertical	2.592	0.200	2.146	3.070
		Injured	Horizontal	2.709	0.253	2.178	4.172
		Uninjured	Horizontal	2.705	0.258	2.256	4.201
Final	Medial	Injured	Vertical	2.499	0.192	2.116	3.071
		Uninjured	Vertical	2.502	0.191	2.148	3.021
		Injured	Horizontal	2.569	0.275	2.112	4.058
		Uninjured	Horizontal	2.562	0.271	2.075	3.890
	Lateral	Injured	Vertical	2.583	0.206	2.148	3.068
		Uninjured	Vertical	2.581	0.189	2.182	3.040
		Injured	Horizontal	2.692	0.270	2.135	4.310
		Uninjured	Horizontal	2.687	0.292	2.198	3.981

NOTE: these are unadjusted means/descriptive statistics. Statistical analysis results shown are for models with sex, age, BMI, and time covariates, and are testing for differences in polynomial fit.

Figure 2.

Lateral Compartment Fractal Dimensions at Baseline and Follow-up. ACL-injured knee (red), ACL-injured contralateral knee (yellow), ipsilateral control knee (green), contralateral control knee (black). Asterisk denotes statistically significant (P<0.05). FD = Fractal Dimension.

Curves correspond to the poly fit of the whole group. *P-vakues are used to determine statiscally significant differences in polynomial fit of the models, and are presented in Table 3.

Figure 3.

Medial Compartment Fractal Dimensions at Baseline and Follow-up. ACL-injured knee (red), ACL-injured contralateral knee (yellow), ipsilateral control knee (green), contralateral control knee (black). Asterisk denotes statistically significant (P<0.05). FD = Fractal Dimension.

Curves correspond to the poly fit of the whole group. *P-vakues are used to determine statiscally significant differences in polynomial fit of the models, and are presented in Table 3.

Table 3.

Whole Group Polynomial Fit Results

Time point	Compartment	FD Variable	P-value
Cases vs Controls: ACL-Injured vs Non-injured Subjects
Baseline	Medial	Vertical	< 0.0001
		Horizontal	0.2
	Lateral	Vertical	0.0009
		Horizontal	0.12
Final	Medial	Vertical	0.01
		Horizontal	0.22
	Lateral	Vertical	< 0.0001
		Horizontal	< 0.0001
Cases: ACL-injured Within Subject Side to Side Comparisons
Baseline	Medial	Vertical	0.87
		Horizontal	0.39
	Lateral	Vertical	< 0.0001
		Horizontal	0.83
Final	Medial	Vertical	< 0.0001
		Horizontal	0.07
	Lateral	Vertical	< 0.0001
		Horizontal	0.003
Controls: Non-injured Subjects Side to Side Comparisons
Baseline	Medial	Vertical	0.06
		Horizontal	0.73
	Lateral	Vertical	0.08
		Horizontal	0.35
Final	Medial	Vertical	0.06
		Horizontal	0.19
	Lateral	Vertical	0.3
		Horizontal	0.09

Data adjusted for sex, BMI age, time from injury to surgery (Baseline data) or time from baseline to final follow-up (Final data).

ACL injured subjects: (Comparison of ACL Injured Knee versus the Contralateral Uninjured Knee within ACL injured subjects)

In the lateral compartment of the tibia (Figure 2), there were significant side-to-side differences in vertical FD for ACL-injured subjects, at both baseline and final follow-up time points (p < 0.0001, for both). The fractal signature on the ACL injured side demonstrated an overall higher FD relative to the contralateral side at baseline and lower overall FD at final follow up. For the horizontal fractal signature in the lateral compartment of the tibia, there was a significant difference in FD profile on the ACL injured side relative to the contralateral side at final follow-up (p = 0.003); however, the same comparison was not significantly different at the baseline visit (p = 0.83).

In the medial compartment of the tibia (Figure 3), there was a significant difference in the vertical FD of the ACL injured limb in comparison to the contralateral side at final follow-up for ACL-injured subjects (p < 0.0001). In contrast, this same comparison was not significantly different at baseline (p = 0.87). For the horizontal fractal signature in the medial compartment, there were no significant differences at both baseline and final follow-up time points (p = 0.39 and p = 0.07, respectively).

ACL-injured versus Control Subject comparison: (Comparison of ACL injured knee with the corresponding healthy knee from the control subject)

In the lateral compartment (Figure 2), there was a significant difference in vertical fractal signature for the ACL-injured subjects’ injured knees in comparison to the corresponding, matched knees of the control subjects at baseline (p = 0.0009), showing an overall higher fractal signature. At final follow-up, the vertical fractal signatures (polynomial fit curves) for the ACL injured subjects injured knee and the corresponding side of the matched control subjects intersected and remained significantly different (p < .0001). Similarly, there was intersection and significant difference of the horizontal fractal signatures between ACL-injured subjects’ injured knees and the corresponding, matched knees of the control subjects, at final follow-up (p = < 0.0001).

In the medial compartment (Figure 3), there were significant differences in the vertical fractal signature for the ACL-injured subjects’ injured knees and the corresponding, matched knees of control subjects, at both baseline and final follow-up (p < 0.0001 and p = 0.01, respectively), showing an overall lower fractal signature. For the horizontal fractal signature, there were no significant differences at baseline or final follow-up (p = 0.2 and p = 0.22, respectively).

Control subjects: (Side-to-side comparison of knees for subjects in the control group)

There were no statistically significant side-to-side differences in vertical or horizontal fractal signature found for control subjects in medial and lateral compartments, at baseline and final follow-up visits (Figures 2 and 3). All of these analyses included adjustment for age, sex, BMI, and time from injury to surgery or time from baseline to follow up.

Discussion

We found significant differences in the fractal signature polynomial curves between ACL-Injured knees and controls with normal knees, as well as side-to-side differences in fractal dimensions within ACL-Injured subjects. To our knowledge, this is the first cohort study with a nested case-control analysis demonstrating subacute changes in tibial subchondral bone trabeculae in subjects at increased risk of developing post-traumatic osteoarthritis using Fractal Signature Analysis in a relatively homogenous group regarding pathology and surgical procedure. Of the 32 participants in the ACL reconstruction group, 28 of 32 (87.5%) ACL reconstructions were performed with autologous bone–patellar tendon–bone (BPTB) autografts, 3 (9.4%) were BPTB allografts, and 1 (3.1%) was reconstructed with a semitendinosus gracilis 4-strand autograft. No patients had more than 1/3 meniscectomy in both medial and lateral compartments combined. Thirteen patients had no tibiofemoral articular cartilage lesions (40.6% of total ACL group), 6 patients had grade 1A or 1B (18.8%), 9 patients had grade 2 lesions (28.1%), and 4 patients had grade 3A lesions (12.5%) as their most significant lesion present.

Our results are similar to those presented by Messent et al(26) in a primary OA cohort, who demonstrated decreases in horizontal and vertical fractal signatures over time. This likely represents early trabecular thickening from sclerosis in the more superficial subchondral bone, however, it may be a mixed picture of both subchondral sclerosis and subarticular osteopenia of the deeper trabecular bone, two separate processes occurring in early PTOA and both resulting in decreased complexity of the image, and therefore lower fractal signatures. Just below the chondral surface, reactive sclerosis may occur in early PTOA leading to redistribution of articular contact stress, reducing load transmission to the underlying subarticular bone resulting in local osteopenia, as discussed by Messent et al.(27). However, the interpretation of fractal signatures is complex. For instance, it is difficult to interpret the relevance of the intersection of the fractal signatures over the radius of coverage between the ACL injured subjects and the uninjured control subjects in the lateral compartment (Figure 2). This may be a limitation of FSA as a tool for evaluating the multiple constructs involved with trabecular bone features as a whole (thickening/thinning of trabeculae, changes in shape, fenestration, etc), as these changes occur at different scale ranges within the “texture” of the bone (i.e., 10–100μm vs. 200–2000μm). Additionally, the use of a relatively large ROI as defined by a commercially-available software package may result in the masking of tibial subchondral trabecular bone changes, which may occur locally within finite regions of each compartment. Future studies evaluating RSA values following ACL injury and surgical reconstruction with other bone-imaging modailities (Dual-energy X-ray Absorptiometry, Computed Tomography, MRI, etc), may prove exceptionally beneficial in the interpretation and validation of results in a group at high risk for sustaining PTOA.(37, 38)

To our knowledge, the previous work of Buckland-Wright et al. is the only other work on fractal signatures in an ACL-injured cohort.(35) Compared to the uninjured knee, the authors found lower horizontal fractal dimensions in the medial compartment. Our results did not reproduce their findings in the medial compartment; however, this may be secondary to differences in the ROI selection(22), as we used a commercially available third-party software package for analyses, and Buckland-Wright, et al., used custom in-house proprietary software. We found no significant difference in the horizontal FD at final follow-up. However, the findings of Buckland-Wright et al. were detected in conservatively treated ACL-deficient knees, four years post-injury when radiographic evidence of osteoarthritis was already present in 50% of injured knees. Unlike the Buckland-Wright et al. study, which found no changes in the lateral compartment, we demonstrate a significantly lower vertical fractal signatures in the lateral compartment when comparting baseline to follow-up.

Roemer et al(39) recently reported a cross-sectional study on fractal signatures of the subchondral tibial plateau of patients (n=685; 135 athletes, 550 non-athletes) seeking medical assessment for subacute or chronic knee symptoms. The study demonstrated a lower horizontal FD among ACL-injured athletes in both the medial and lateral compartments compared to the rest of the cohort. Our results support their findings as we also demonstrated a lower FD in the lateral compartment at final follow-up. Further, earlier results(40) from the same cohort found prior ACL surgery in these young athletes to increase the odds of having radiographic evidence of OA compared to non-athletes at follow up, suggesting there may be a relationship between the fractal signatures and the development of OA.

Other longitudinal studies have examined tibial subchondral bone in the context of primary OA using FSA.(21, 26, 41, 42) Most recently, Kraus et al(21) reported lower fractal signatures of horizontal trabeculae in patients with progressive OA compared to nonprogressive OA, suggesting that changes in periarticular bone are sensitive indicators for OA disease progression. We utilized a similar quadratic polynomial fit to assess fractal signature changes over time and also found lower horizontal fractal signatures in the lateral compartment of ACL-injured knees. Unlike out results, Kraus et al found higher vertical fractal signatures. It is possible that the subchondral thickening associated with primary idiopathic OA shares a similar pathology early in the time course of PTOA. It should also be noted that other groups have found similar results in larger primary OA cohorts(22–24), however, different analysis methodologies and evaluated trabecular scales (μm ranges) may explain subtle differences in results.

Lastly, we have also reported the first side-to-side comparison of fractal signatures in healthy subjects with normal knees. With no significant side-to-side differences, these findings provide a supporting rationale for using the healthy contralateral knee as a control for the injured knee in future studies.

Potential limitations of this investigation include a small sample size and only midterm follow up, making it impossible to correlate our findings with the incidence of radiographic OA. However, these limitations are offset by important strengths that include the prospective study design, the focus on a homogenous group of subjects that suffered their first ACL trauma with minimal associated injury to other articular structures such as the articular cartilage and menisci, evaluation of healthy control subjects with normal knees, and the use of standardized fluoroscopy-assisted radiographs as recommended by Jarrraya 2015(43) represent important strengths. It should also be noted that the p-values shown herein are measures of statistical significance. The p-value is actually the probability of obtaining an estimate, or test statistic, at least as extreme as the one obtained, if the null hypothesis for the statistical test were true. Thus it is a measure of inferential uncertainty, as opposed to clinical importance. The statistically significant results shown may not indicate clinical or practical importance in all cases. In addition, the use of 3T MRI at baseline and follow-up of both ACL injured and control subjects (to document they had normal knees at baseline and follow up), represent important strengths.

In conclusion, we found significantly lower fractal signatures for ACL-injured knees over time, when compared to the contralateral healthy knee as well as uninjured controls. These results may suggest thickening of the subchondral bone following ACL injury and reconstruction at medium-term follow up. The changes in the trabecular structure following ACL-injury reported here provides an opportunity for further investigation and possible intervention during the earliest, asymptomatic stages of PTOA. FSA is a resilient, cost-effective method to evaluate boney microarchitecture during the onset and progression of PTOA. However, analyzing and interpreting the initial changes in fractal dimensions soon after joint trauma and how they are related to PTOA will require assessments at longer term follow-up intervals, and additional work is required to standardize fractal signature data in order to translate our findings to other study populations.

Figure 1.

FSA Region of Interest. Serial fractal signatures of tibial subchondral bone architecture were calculated in medial (blue) and lateral (red) regions of interest. The tibial joint profile was defined using a 6-point manual initialization (medial and lateral femur, medial and lateral tibial, and tibial spines).