1. Introduction
Anterior cruciate ligament (ACL) rupture causes knee instability, which affects sport activities and increases the risk of meniscal injury and early osteoarthrosis [
1]. ACL reconstruction reduces anterior tibial laxity but does not fully restore normal tibiofemoral kinematics [
2]. Despite efforts and significant advances in ACL reconstruction, clinical failures continue to occur [
3]. It is still debated among researchers and clinicians on what processes are involved in ACL graft healing and remodeling and how long these processes take in patients [
4]. This information is crucial for patients to return to sports activities.
Claes et al. [
5] performed a systematic review of literature on the “ligamentization” process of human ACL graft. He showed that a free tendon graft used for ACL reconstruction undergoes a series of biologic processes called “ligamentization”. Ligamentization may be divided further into the early phase, remodeling, and maturation of the graft. The graft shows viability throughout the whole process. In histopathologic evaluation, the graft resembles a normal intact human ACL. However, specific differences exist between the phases. In addition, there is no consensus for the exact time frames for each stage of the ligamentization process.
Biercevicz et al. [
6] proved that volume and greyscale values are predictive of the healing graft’s structural properties in his preclinical animal model studies. He also showed magnetic resonance imaging (MRI) parameters may predict biomechanical properties and outcome measurements in human patients after ACL reconstruction [
7]. Weiler et al. [
8] also showed in an animal model that quantitative evaluation of MRI might be a useful tool for following the graft remodeling. Dong et al. [
4] proved that implanted grafts could transform into native ACL-like tissue with a similar ultrastructure and metabolism.
The aim of our study was to evaluate the bone-patellar tendon-bone (BPTB) ACL graft remodeling process in our group of patients. We used previously described techniques to evaluate their efficacy. Our results were correlated with variables including age, sex, and weight-bearing. Additionally, for the first time to our knowledge, we correlated the graft remodeling process not only with clinical outcome, but also with indirect MRI signs of ACL graft insufficiency. We also analyzed standard deviation of SI as a separate value.
2. Materials and Methods
Our retrospective analysis included 44 patients who underwent BPTB ACL reconstruction in our hospital between December 2011 and March 2015. There were 23 males and 21 females. The mean age of patients was 28 years old (range 13–73). There were 18 right and 26 left knees. At the time of final evaluation (96 weeks postoperatively), all patients, except one, were stable on clinical evaluation. This patient with ACL graft complete injury had a traumatic secondary graft failure.
The rehabilitation protocol was the same for all patients. The only difference was weight-bearing immediately after the surgery, and this factor was also evaluated separately in this study. One group of patients had full weight-bearing immediately after the surgery. The second group had no weight bearing for the first three weeks after the surgery. A brief description of the rehabilitation protocol is presented in
Table 1.
MRI was performed in all patients postoperatively at 2, 6, 12, 18, 24, 36, 48, 72, and 96 weeks. MRI images were obtained using a 1.5 tesla GE Signa HDxt scanner with an 8-channel knee coil. Full, routine MRI protocol for ACL assessment included axial and sagittal PDWI (Proton Density Weighted Images), axial FSE STIR (Fast Spin Echo Short Tau Inversion Recovery), DWI (Diffusion Weighted Images), and T2 Mapping sequences.
Signal intensity was measured manually using the regions of interest (ROI) tool with a diameter of 3 mm on axial and sagittal PDWI sequences (
Table 2).
To measure ACL graft signal intensity, we used the method previously described by Howell et al. [
9], Ahn et al. [
10] and Vogl et al. [
11], by calculating signal-to-noise quotient (SNQ, Equation (1)):
We wanted to evaluate which measurements (sagittal or axial plane) are more reliable for predicting ACL graft insufficiency. Therefore, we calculated separately for both planes.
We also evaluated age, sex, immediate postoperative weight-bearing, and intraoperative tensioning of the graft, which is the initial position of the tibia in relation to the femur.
The lower value of SNQ—the better graft appearance, the better mechanical properties of the graft and the more comparable ACL graft is to posterior cruciate ligament (PCL) healthy ligament.
2.1. Measurement on the Sagittal Plane
A sagittal plane in the PDWI sequence with the best visibility of the ACL graft was chosen for evaluation. We manually drew ROI (diameter of 3 mm) in 1/3 distal and 1/3 midsubstance of the graft graft, PCL, and the background (1 cm distal from the tip of the patella and 1 cm anterior to the skin) (
Figure 1). The SNQ for this plane was calculated by averaging SNQs calculated separately from the two measurements.
2.2. Measurement on the Axial Plane
The lowest most distal level of the intercondylar notch was chosen for measuring the axial plane. ROI (diameter of 3 mm) was drawn on ACL graft, PCL, and background (1 cm anterior to the skin on the knee’s anterior aspect) (
Figure 2). SNQ was calculated using Equation (1).
2.3. Standard Deviation (SD)
We evaluated the standard deviation (SD) of each measurement as a separate value. Since SD represents average difference from the mean value, lower SD indicates that the values within the ROI are more similar to the mean value, thus the graft is more homogenous. Similarly, higher SD indicates more heterogenous graft. To incorporate standard deviations of the ACL graft signal intensity, we computed relative SD index (rSD, Equation (2)):
For the measurements from the sagittal plane, a sum of the two measurements was used in the numerator.
2.4. Indirect Signs of Anterior Cruciate Ligament (ACL) Graft Insufficiency
We indirectly evaluated ACL graft insufficiency and correlated our findings with other parameters. Parameters taken into consideration were:
- −
ACL graft appearance: normal/injured (
Figure 3a,b)
- −
PCL appearance: normal buckling (
Figure 4a,b)
- −
Measurement of anterior tibial subluxation was made in the sagittal plane at a location midway between the cortex adjacent to PCL in the femoral notch and the most lateral slice containing the lateral femoral condyle. In
Figure 5, two vertical lines were drawn tangential to the posterior cortical margin of the lateral femoral condyle and the lateral tibial condyle. Both lines were drawn parallel to the margin of the image frame. The degree of tibial subluxation was measured by the distance between the two lines in millimeters. Anterior tibia subluxation was evaluated as (+), and posterior tibia subluxation (in relation to femur position) as (−).
2.5. Data Analysis
Statistical Analyses
We used R 4.0.2 [
12] for performing statistical analyses. To analyze changes in SNQs and indirect signs of ACL graft insufficiency over postoperative timepoints and their relationships to rSDs, sex, age, and weight bearing, a series of Bayesian multilevel regression was conducted with intercepts allowed to vary by patient. Details of the models are provided with the results. All continuous variables were entered into the models after transformation to Z-scale and categorical predictors were coded with orthogonal sum to zero contrasts. Change over weeks (week predictor) was modelled with orthogonal quadratic polynomial, which allowed us to capture non-linear trend in change of the parameters. For all categorical predictors, an interaction with week predictor was also investigated to test whether the change over weeks depended on these predictors. Each model was estimated separately for measurement made in sagittal and axial planes.
In Bayesian statistics, the inference is based on analyzing posterior probability distributions of model parameters (e.g., regression weight of week predictor), obtained by integrating likelihood with prior probability distributions of the model parameters. The model parameter is statistically credible when 95% of credible intervals (CI) of the posterior distribution exclude zero [
13]. As a point estimate of the effect, medians of the posterior distributions are presented. As priors for all regression weights, non-informative, standard normal distributions were used (i.e., with M = 0 and SD = 1).
To present expected values of clinical parameters, we provide posterior predicted marginal means for SNQs and PCL subluxation and probabilities for overall graft state and PCL anterior subluxation. These values were computed from model parameters and can be thought of as better summary statistics than simple descriptive statistics, since they incorporate all the uncertainty in the data and model.
For models with continuous dependent variables, Bayesian R2 [
14] was provided as a measure of model fit. Therefore, prediction accuracy was reported for models with categorical dependent variables. Both statistics are in the range zero to one, with one indicating perfect model fit.
To approximate posterior distributions of the models, a Markov Chain Monte Carlo (MCMC) sampling procedure was conducted using the brms (Bayesian Regression Models using ‘Stan’) package [
15]. For each reported model, six parallel MCMC chains were used. Each chain consisted of 6000 samples, with 3000 samples used as warmup period and every 10th sample recorded, which resulted in a total of 1800 recorded samples. Efficient sampling procedures resulted in well-mixed and autocorrelation-free chains and unimodal posteriors.
4. Discussion
To the best of our knowledge, this is the first complex study evaluating ACL graft maturity by comparing different methods for measuring graft signal intensity, homogeneity of the graft, and evaluating the effect of sex, age, and weight-bearing being believed to have an impact on graft maturity. Moreover, we studied indirect MRI’s signs of ACL graft insufficiency and its relationship to signal intensity.
In 2019 Van Dyck et al. [
16] performed a systematic review of papers assessing ACL graft maturation in MRI studies. He concluded that methods used across studies were too heterogeneous to conclude time frames of signal intensity. The signal intensity of the graft was poorly correlated with clinical outcomes.
Van Groningen et al. [
17] published a systematic review of papers assessing ACL graft maturity. He identified 10 studies with serial MRI measurements of SNQ including 479 patients (20–98 patients per study). Semi-tendinous and gracilis tendons were used for ACL reconstruction (six studies), quadriceps tendon (one study), and bone-patellar tendon-bone grafts (3 studies). Allografts were used in three studies, apart from autografts. In all studies, SNQ was calculated either by comparing signal intensity of the graft to the posterior cruciate ligament or the background. Based on available data, Van Groningen concluded that MRI SNQ was the highest at six months postoperatively and later had a gradual decrease over time. However, in his opinion, the MRI SNQ method should not be used to predict graft maturity and functional and clinical outcomes.
In our study we verified different methods for evaluating ACL graft maturity: ROI and SNQ methods on axial and sagittal plane as well as rSD values. It is the first study to compare measurements obtained in both sagittal and axial planes. We showed that average values from two measurements of ACL graft taken in the sagittal plane best represent graft status. We also determined the SD of signal intensity of the graft as a new parameter for graft evaluation and prediction of graft insufficiency. The rSD was related to SNQ and overall grafts state, indicating that graft homogeneity is also an important indicator of graft well-being. We also evaluated the relationship between graft maturity and more subtle signs of ACL graft insufficiency, such as graft status, PCL buckling and exact measurement of anterior tibia subluxation.
Fukuda et al. [
18] evaluated SNQ for ACL graft maturation in 10 multisequence MRIs performed within 50 months postoperatively. Forty-five patients with a mean age of 27 years old were included in the prospective study. All patients underwent double-bundle ACL reconstruction. Subsequently, all patients received a MRI at 3 weeks and 6, 9, 12, 18, 24, 36, 48, and 50 months after surgery. In the conclusion of the study, Fukuda stated that at least 18 months was needed for the mean SNQ of the anteromedial bundle for normalization, whereas for the postero-lateral bundle, at least 24 months was needed.
Zaffagnini et al. [
19] showed that even though hamstring tendons used for ACL reconstruction undergo a transformation, they do not match the ultrastructure pattern of a native ACL for up to ten years.
In our study, we showed that even at 96 weeks postoperatively, signal intensity of the graft did not match values measured immediately after the reconstruction (at two weeks). Our data suggest that normalization of the signal make take even longer. The question “what would be the best time for patients to return to full sport activity” remains unanswered.
We did not find any significant association between graft maturity and age of the patient at the time of surgery. However, further studies including more heterogeneous groups are needed, as our group of patients was rather young (mean age 28 years old). Sex of the patient did not have an effect on graft maturity. On the other hand, full weight bearing immediately after surgery seemed to improve graft remodeling with smaller values of signal intensity. This finding is in accordance with a study performed by Dziki et al. [
20] showing that appropriate mechanical loading improves healing and remodeling process.
Van der List et al. [
21] demonstrated that postoperative MRI of the ACL graft might accurately predict the risk of re-rupture of the ACL graft. Our study also presents a good correlation between measured SNQ and secondary symptoms of graft insufficiency.
A particularly interesting finding of this study is that regardless of initial graft tension at the time of surgery, in order to obtain proper positioning of the tibia relative to the femur, the trend of subsequent anterior tibia subluxation remains comparable. Specifically, patients with over-tension (starting with posterior tibia subluxation) end up with smaller degrees of anterior tibia subluxation at 96 weeks. Currently, pre-operative rehabilitation for chronic patients aimed at reducing anterior tibia subluxation might play a key role for final outcome, facilitating proper positioning of the tibia versus femur intraoperatively.
The question regarding the best time to come back to full sports activity remains unanswered, and the literature is conflicting at that point.
Zaffagnini et al. [
22] concluded that the best situation would be patient-tailored rehabilitation protocols and return to sport criteria based on individual characteristics. On the other hand, SIGASCOT (Italian Society of Knee, Arthroscopy, Sport, Cartilage and Orthopaedic Technologies) published with Grassi et al. [
23] data reflecting on return to sport after ACL reconstruction. 123 SIGASCOT members took part in 14 questions survey focusing, among other things, on return to sport criteria. Returning to non-contact sports was allowed within six months by 87% of members and 53% for contact sports. Return to competitive sport was allowed after six months for 48% for non-contact sports and only 13% for contact sports. The most used criteria to qualify for return to sport were a full range of motion (77%), Lachman test (65%), and Pivot-Shift test (65%). MRI criteria were used only by 12% of surgeons. Nagelli et al. [
24] pointed out strong evidence indicating that nearly one-third of younger patients will sustain a second ACL injury within the first two years after ACL reconstruction. He postulates that delaying return to sport for two years after ACL reconstruction will significantly reduce secondary ACL graft injuries. According to Culvenor et al. [
25], an accelerated (<10 months) return to sport after ACL reconstruction may be implicated in developing knee osteoarthritis development.
In our opinion, return to sport after ACL reconstruction should individually be assessed based on several factors, where MRI is only one of those:
- −
Psychological readiness for sport;
- −
No pain;
- −
Clinically stable knee (Lachman test, Rollimetr/KT 1000);
- −
Minimum ROM 0–130;
- −
MRI—good signal of the graft;
- −
Biomechanical evaluation: symmetrical muscle strength (on the level of pelvis and knee), good proprioception, movement analysis of simple jump and landing (always paying attention to pelvic stability and valgus knee position). If it is not possible to do an evaluation in the professional lab, a simple, functional test and video analysis based on a cell-phone would also work.
Conflicting data in the literature regarding MRI ACL graft evaluation probably come from the fact that authors tried to use this as a single factor for return to sport assessment or were correlating it with clinical outcome. Return to sport criteria must be multifactorial and lack of correlation of MRI with clinical stability is a positive thing as it may show us more than we may detect with clinical evaluation.