This is the most comprehensive multivariable analysis of a prospective multicenter cohort of sufficient size and rate of follow-up to demonstrate that variables measured at the time of ACLR (revision ACLR, allograft, lateral meniscus status, BMI, smoking status) are predictors of six-year sports activity and function as measured by the IKDC, KOOS, and Marx activity level outcome instruments. Each of these predictors (variables) is modifiable except for revision ACLR. Thus, avoiding allograft as a graft choice, leaving “stable” partial and complete lateral meniscal tears alone, not smoking, and maintaining a relatively lower BMI could improve ACLR outcomes. In contrast to the modifiable predictors for IKDC and KOOS, the predictors of declining Marx activity (revision ACLR, and female sex) are not modifiable. However, despite the decline in activity level, the population medians of the cohort remain at the same two-year IKDC and KOOS subscale levels.
The maintenance of IKDC and KOOS outcomes at six years was an unexpected result. We anticipated a decline from the two-year outcomes in all three scales, which were clearly not observed. These results indicate that our present technique of ACLR is durable at the six-year mark. The potential role of declining Marx activity level to reducing knee-related stress and therefore preserving joint health, as would be measured in the future by IKDC and KOOS, is unknown. While it may take more time before declination of knee function is observed in this cohort, the similar group score at two and six years for the validated, patient-reported outcomes provide a good prognosis to be conveyed to our patients preoperatively.
A comprehensive systematic review by Oiestad et al
. evaluated knee OA after ACL reconstruction and found that concomitant meniscus tears were associated with radiographic OA using univariate analysis. 35
Unfortunately, the authors were unable to perform a meta-analysis due to the heterogeneous classification systems defining OA, the lack of inter-rater agreement, and lack of multivariable analysis. 35
They concluded that future studies that define both the prognosis and predictors of OA after ACL reconstruction should be prospective with clearly defined aims and endpoints, include clear inclusion and exclusion criteria, utilize a common radiographic classification system with reliability data and independent blinded examiner, the rehabilitation protocol should be reported, and that regression analysis be used to evaluate risk factors.35
We believe that the majority of these points characterize the current cohort. The strengths of this study include the application of multicenter prospective longitudinal assessment utilizing the same validated outcome measures over time and accruing greater than 85% follow-up, which is the preferred research design (level I) to evaluate prognosis and modifiable predictors through multivariable analysis.35
In clinical practice, patients have many different combinations of potential predictors that can be independently scaled and then summed to yield a patient-specific result. This result can be obtained through use of an equation where individual values are entered or by the use of a nomogram. Patients present with an almost infinite combination of these variables and an individual’s specific outcome now can be estimated. Alternatively, an individual surgeon can avoid allograft ACLR graft, counsel patients on smoking cessation and maintaining healthy weight, and leave stable lateral meniscal tears alone in an effort to improve the outcomes of his or her patients. The multicenter nature of this consortium lends the results to be generalizable to patients treated by fellowship-trained sports medicine physicians.
There are several weaknesses in this analysis. Despite being the largest prospective cohort utilizing multivariable analysis for ACLR outcomes, our sample size is still too small to provide a more detailed analysis to the injuries involving the articular cartilage and meniscus. Due to the relatively low frequency of chondromalacia grades II, III, and IV, these are grouped together in the current analysis. Ideally, as additional subjects are prospectively enrolled and evaluated at six years, our modeling can be divided into more clinically applicable chondromalacia grades (II vs. III vs. IV). Previous inter-rater agreement27
has shown our ability to divide by individual grade. Likewise, meniscus excisions are currently all grouped together instead of by stratifying by one-third, two-thirds, or whole which has greater clinical meaning. Another weakness is the lack of important complimentary information gathered by clinician observation and testing of knee joint laxity, physical characteristics, and radiologic images of the ACL reconstructed knee. The logistical and financial requirements of onsite follow-up impede performing sufficiently powered multivariable analysis on equally important patient-reported outcomes (such as sports function, knee quality of life, and activity level) specifically designed to follow much larger cohorts. However, information regarding the principal outcomes that influence a surgeon’s and patient’s decision making -- clinical failure, restoration of functional stability, activity level and sports participation or function, pain, reoperation, and function in activities of daily living (ADL) can be gathered through the use of validated questionnaires and patient interview.
Several prospective and retrospective studies have explored predictors or risk factors for ACLR through a variety of statistical methods. Recently a randomized controlled trial (RCT) at ten-year follow-up showed no difference between autograft hamstring and patellar tendon in clinical assessment (laxity, hop, isokinetic strength), radiographic osteoarthritis or patient-reported outcomes (Cincinnati and Lysholm).11
Likewise an RCT between neuromuscular versus traditional strength rehabilitation did not demonstrate a difference for Cincinnati or Lysholm at 2 years.39
Similarly several studies failed to demonstrate a correlation between clinical assessments and validated patient-reported outcomes (KOOS, SF-36, IKDC).21, 23, 33, 49
However, two studies found several clinical assessments significantly affected ACLR outcome.38, 51
Decreased range of motion in knee extension, meniscectomy, presence of articular cartilage damage, and time from injury to surgery all led to significantly worse IKDC outcomes and radiologic OA.51
However, a 7-10 year longitudinal cohort study on both patellar tendon and hamstring tendon ACLR did not find extension range of motion as a risk factor for radiologic OA.38, 41
However, they did observe that patients undergoing a patellar tendon autograft ACLR had more radiologic OA. Analogous to our predictors of BMI and smoking, several prior studies have likewise demonstrated they are risk factors for patient-reported outcomes.17, 22, 52
Also in agreement with our finding, age and gender were not risk factors for patient-reported outcomes.33, 41
In our multivariable analysis, education level, prior meniscectomy, and medial meniscus status were not risk factors which have been previously shown by others.6, 22, 23, 49, 52, 54
In addition, other factors not explored in our model that have been shown to be risk factors include preoperative quadriceps strength,6
knee self-efficacy scale (KSES),53
and patient satisfaction.20
We believe our study was underpowered to test the effect of meniscus and/or articular injury and treatment with a single year’s cohort. When a second year is followed, we expect adequate sample size to evaluate.
The multivariable analysis most similar to our analysis found a pop at injury (KOOS, IKDC, Lysholm), no change in educational level (KOOS and IKDC), and weight gain greater than 15 pounds (IKDC) to be predictors of their respective outcomes.52
Our analysis found high baseline BMI to be predictive of poorer IKDC and KOOSsports/rec
subscale scores. However, just as in the prior study, the individual differences in outcome were below a clinically meaningful difference for each outcome measure. We did not find a pop heard at the time of injury to be significant in any outcome. The prior study did not evaluate smoking, allograft, or revisions, and the current study did not evaluate educational level. However, both studies found age and gender were not related to outcomes.
The major role that revision vs. primary ACLR has on every outcome measure clearly supports the role of additional research aimed at understanding and improving outcomes after revision ACLR. Thus, the importance of a multicenter study of revision ACL reconstructions is once again confirmed. Multivariable analysis of a large group of revision ACL reconstructions will be necessary to determine the predictors in revision surgery for these poor outcomes. Since revision ACLR has such a large negative effect on outcome even when controlling through multivariable analysis for articular cartilage and meniscus injuries and treatment, every effort should be made for secondary prevention of ACLR graft tear.
In conclusion, our MOON results found that choosing an autograft would significantly and in a clinically meaningful way improve sports function (IKDC, KOOSsports/rec) and knee-related quality of life (KOOSkrqol), whilst not smoking is associated with better IKDC and KOOSkrqol scores, and a lower BMI is predictive of better IKDC and KOOSsports/rec scores. The actual improvement in outcomes can be predicted for each outcome by use of the respective nomograms. Unfortunately, no modifiable predictors were identified for the declining Marx Activity scale. Since revision ACLR has the most powerful negative effect on outcome, secondary prevention strategies should be explored and tested.