Background Anterior cruciate ligament (ACL) injury is an important risk factor for development of knee osteoarthritis (OA). To identify those ACL injured patients at increased risk for knee OA, it is necessary to understand risk factors for OA.
Aim To summarise the evidence for determinants of (1) tibiofemoral OA and (2) patellofemoral OA in ACL injured patients.
Methods MEDLINE, EMBASE, Web of Science and CINAHL databases were searched up to 20 December 2013. Additionally, reference lists of eligible studies were manually and independently screened by two reviewers. 2348 studies were assessed for the following main inclusion criteria: ≥20 patients; ACL injured patients treated operatively or non-operatively; reporting OA as outcome; description of relationship between OA outcome and determinants; and a follow-up period ≥2 years. Two reviewers extracted the data, assessed the risk of bias and performed a best-evidence synthesis.
Results Sixty-four publications were included and assessed for quality. Two studies were classified as low risk of bias. Medial meniscal injury/meniscectomy showed moderate evidence for influencing OA development (tibiofemoral OA and compartment unspecified). Lateral meniscal injury/meniscectomy showed moderate evidence for no relationship (compartment unspecified), as did time between injury and reconstruction (tibiofemoral and patellofemoral OA).
Conclusions Medial meniscal injury/meniscectomy after ACL rupture increased the risk of OA development. In contrast, it seems that lateral meniscal injury/meniscectomy has no relationship with OA development. Our results suggest that time between injury and reconstruction does not influence patellofemoral and tibiofemoral OA development. Many determinants showed conflicting and limited evidence and no determinant showed strong evidence.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Anterior cruciate ligament (ACL) rupture is a common sports-related injury, with an annual incidence of approximately 5/10 000 persons in the general population.1 Osteoarthritis (OA) is a well-known, long-term complication of ACL rupture, with a prevalence of 10–90% at 10–20 years postinjury.2 ,3 It is important to identify the risk factors contributing to OA in patients with ACL rupture, because some risk factors may be modifiable as to prevent onset or early-stage progression of OA. At present, the only treatment options for OA are symptomatic relief, osteotomy, unicompartmental arthroplasty and, for end-stage disease, total knee arthroplasty. Early intervention is critical because patients with post-traumatic OA are typically young and it is important to postpone total knee arthroplasty.4
Numerous studies have evaluated the long-term consequences of ACL rupture. These studies are heterogeneous with regard to methodology, including treatments used, inclusion of additional intra-articular injuries, reported OA outcomes and descriptions of determinants (potential risk factors). Three previous systematic reviews of development of OA after ACL rupture were limited either because they considered OA only in the tibiofemoral compartment or because they focused on one type of treatment (ACL reconstruction). Oiestad et al5 conducted a systematic review of the prevalence of OA in the tibiofemoral joint occurring more than 10 years after ACL injury. They included studies that used ACL reconstruction techniques, which are no longer used (eg, Leeds-Keio polyester ligament surgery or suturing of the ACL). Therefore, we did not include these techniques in this systematic review. To better evaluate newer and current techniques and rehabilitation methods, we included only studies which reported results based on current ACL reconstruction procedures. Magnussen and Spindler6 reviewed patient factors affecting clinical and radiographic outcomes after ACL reconstruction in prospective studies with a 5-year minimum follow-up. Prospective study design was an inclusion criterion, so they missed the results of all retrospective studies. Claes et al7 reviewed the literature on long-term radiographic outcome after autologous ACL reconstruction; studies with a mean follow-up of less than 10 years were excluded. They investigated only one predictor, namely the relationship between meniscal status and OA development in the reconstructed knee. Currently, there is no consensus about operative or non-operative treatment for preventing OA, and degenerative changes can develop in all knee compartments.
Culvenor et al8 showed in their narrative literature review that patellofemoral OA after ACL reconstruction occurs as frequently as tibiofemoral OA. Different mechanisms, such as inflammation, concomitant injuries to the patellofemoral articular cartilage, meniscal injury, graft choice and changes of knee biomechanics, may play a role in the development of patellofemoral OA.8
The previous published reviews presented a part of the general question: Which determinants influence the development of degenerative changes after an ACL rupture? This systematic review will fill the gaps of the previous reviews and supplement recent published literature on both tibiofemoral and patellofemoral OA. We systematically reviewed the evidence for determinants of both (1) tibiofemoral OA and (2) patellofemoral OA in patients with an ACL injury treated operatively or non-operatively.
The reporting in this systematic review was conducted according to the PRISMA statement.9
Data sources and searches
MEDLINE, EMBASE, Web of Science and CINAHL medical literature databases were searched up to 20 December 2013. Search terms included anterior cruciate ligament, synonyms for injury and synonyms for osteoarthritis. The full electronic search strategy for the MEDLINE database is presented in table 1. Similar search strategies were used in EMBASE, Web of Science and CINAHL. Additionally, the reference lists of all eligible studies were manually screened.
Two reviewers (BLvM and MR) assessed the studies for the following inclusion criteria:
The following study designs with at least 20 patients: randomised controlled trial, prospective follow-up study, matched case–control study and retrospective study;
Subjects had to have an ACL injury consisting of:
Patients treated non-operatively or
Patients treated operatively; use of an arthroscopic or miniarthrotomy technique and use of bone-patellar tendon-bone, hamstring tendon or allografts;
Written in English, German, Dutch, Spanish, French, Swedish, Danish or Norwegian;
Full text available;
Measured one of the following OA outcomes:
Clinical OA: according to a clinician, self-reported or American College of Rheumatology (ACR) criteria10; osteotomy, unilateral knee arthroplasty or total knee arthroplasty (indirect measures for clinical knee OA);
OA findings on MRI;
OA findings during arthroscopy;
The relationship between outcome and determinant, defined as potential risk factors, must have been described or data must be available to calculate the relationship;
Determinant studied in ≥2 studies;
Determinant must be measured prior to the OA outcome;
Follow-up period of at least 2 years.
Animal studies and reviews were excluded. Disagreements on inclusions were resolved by discussion and, if necessary, a final decision was made by a third reviewer (JANV).
Data extraction and risk of bias assessment
Two reviewers (WAvE and BLvM) extracted the study characteristics, follow-up times, determinants, outcomes and the relationship between outcome and determinant.
The determinants were grouped into patient characteristics (age, body mass index (BMI), sex), physical examination, activity level and intra-articular-related factors. The determinant that was named laxity consisted of results of a pivot shift test, Lachman test, KT 1000 arthrometer or description of ‘laxity’. The location of injury of the intra-articular determinants: chondral injury and meniscal injury/meniscectomy were presented when reported as such in the studies. For determining the influence of tunnel placement on OA development, we used the assessment of tunnel position when a study evaluated both femoral and tibial tunnel positions and graft inclination. If studies had the same population and determinant, but different follow-up times, we presented the results of the study with the longest follow-up time. When a determinant was measured in various ways and had different relationships with OA outcome in one study, all results were presented. For the analyses of the relationship between determinants and OA outcome, the distinction between patellofemoral and tibiofemoral OA was made. If the studies did not report a specific compartment for the OA outcome or if the studies reported the OA outcome for all compartments, then the study was classified as OA outcome in which the compartment was unspecified. Since the included studies presented the relationship between determinant and OA outcome in various ways, we reported the presence of a ‘positive significant relationship’ or ‘negative significant relationship’ or ‘no significant relationship’. For presentation of the results, we distinguished the studies into two groups: (1) studies with inclusion of non-operatively treated patients and (2) studies with inclusion of both operatively and non-operatively or solely operatively treated patients.
We evaluated the selected studies on 12 aspects using modified questions of existing risk of bias assessment tools.11–13 Our assessment tool contained questions about the aim of the study, description of inclusion and exclusion criteria, collection of data, validity and reliability of OA outcome measures, independent measure of determinants, valid and reliable measurement of determinants, follow-up period, loss to follow-up, and use of adequate statistical analyses. Four reviewers independently assessed the quality of the included studies. Disagreements were resolved by discussion. Studies were classified as low risk of bias when they scored ‘adequate’ on all the following topics: the authors reported inclusion of consecutive patients; there was unbiased assessment of the study outcome and determinants; the determinant measures were used accurately (valid and reliable); if there was a loss to follow-up of less than 20% and there was a description of the reasons, and if there was correction for confounding. The assessment tool used is given in online supplementary appendix table S1.
Data synthesis and analysis
Since the studies were considered clinically heterogeneous with regard to the outcome measures and determinants studied, it was not possible to pool the data for statistical analysis, and therefore we performed ‘a best-evidence synthesis’.14 ,15 With the use of the system developed by van Tulder et al,16 the following ranking of levels of evidence was formulated: (1) Strong evidence is provided by two or more studies with low risk of bias and by generally consistent findings in all studies (≥75% of the studies reported consistent findings). (2) Moderate evidence is provided by one low risk of bias study and two or more high risk of bias studies and by generally consistent findings in all studies (≥75%). (3) Limited evidence is provided by one or more high risk of bias studies or one low risk of bias study and by generally consistent findings (≥75%). (4) Conflicting evidence is provided by conflicting findings (<75% of the studies reported consistent findings). (5) No evidence is provided when no studies could be found.16 ,17
Identification and selection of the literature
The search resulted in 2348 studies, for which all abstracts were reviewed. After screening of the abstracts, 157 were identified as possibly relevant, and full texts were retrieved. After review of the full texts, 56 met all the inclusion criteria (figure 1). There were no disagreements on inclusions. The references of all 56 studies were reviewed and 8 additional studies meeting the inclusion criteria were identified. Thus, 64 studies in total were included in this systematic review.
Description of the included studies
The characteristics of the included studies are presented in online supplementary appendix table S2. The studies had the following designs: randomised controlled trial (n=12),18–29 prospective follow-up study (n=22),30–51 matched case–control study (n=2)52 ,53 and retrospective study (n=28).54–81 The number of patients available for follow-up measurement in the studies ranged between 30 and 780. In 62 studies, the OA outcome was determined with radiographs, and in 2 studies by MRI assessment.28 ,47 Only two studies43 ,70 reported radiological OA and clinical OA as outcomes. Therefore, the findings of this systematic review address the influence of radiological OA. In 47 studies (4956 patients), the treatment strategy was ACL reconstruction, in 4 studies22 ,64 ,71 ,76 (273 patients) non-operative treatment, and in 13 studies19 ,30 ,31 ,40 ,41 ,47 ,53 ,65 ,70 ,72 ,77 ,79 ,80 (1169 patients) both reconstruction and non-operative treatment. The mean follow-up time varied between 3.9 and 20 years.
Risk of bias assessment
Two studies35 ,56 were classified as ‘low-risk of bias’. An overview of the quality assessment score of the included studies is presented in supplementary appendix table S3. The main aim of the two low risk of bias studies was to investigate risk factors for development of knee OA after ACL reconstruction. In these studies, the number of patients used for analyses was >50; Ahn et al had a sample size of more than 100 patients (n=117). Janssen et al used only hamstring tendon grafts, and Ahn et al bone-patellar tendon-bone grafts.
Influence of determinants in non-operatively treated patients
Four studies22 ,64 ,71 ,76 included solely non-operatively treated patients. Limited evidence was found for a positive relationship between meniscectomy and development of knee OA in chronic ACL-deficient knees. Determinants age, BMI and sex were excluded because they were studied in only one study. The influence of laxity on OA development could not be presented because it was measured concurrently with the OA outcome.
Influence of determinants in both operatively and non-operatively or solely operatively treated patients
Conflicting evidence was found for the influence of age on OA outcome in all compartments (tables 2⇓–4). For the influence of BMI on OA outcome in the tibiofemoral compartment and compartment unspecified, conflicting evidence was found after ACL rupture. Limited evidence for no relationship was found for OA development in the patellofemoral compartment after ACL rupture. Nine studies evaluated the relationship between sex and OA development after ACL rupture. For development of tibiofemoral OA, three high risk of bias studies43 ,60 ,66 showed conflicting evidence. Moderate evidence was found for no relationship between male sex and OA development in compartment unspecified.25 ,35 ,67 ,68 ,75 ,79
One low risk of bias35 and two high risk of bias34 ,45 studies showed no relationship between laxity and development of OA in compartment unspecified (table 4). Thus, there is moderate evidence for no relationship between laxity and OA development.34 ,35 ,45 Moderate evidence was also found for no relationship between range of motion and OA development in compartment unspecified.34 ,35 ,45 ,50 Performance of a single-legged hop test was evaluated in three studies34 ,35 ,45 and showed conflicting evidence.
One low risk of bias study35 and one high risk68 of bias study found no significant relationship between activity level before reconstruction and OA development (compartment unspecified; table 4).
Two high risk of bias studies44 ,63 investigating additional injuries in general showed conflicting evidence (tables 2⇑–4).
One high risk of bias study68 evaluated patellar, medial and lateral chondral injury after ACL rupture and their influence on OA development in compartment unspecified. Medial and patellar chondral injury showed a positive significant relationship with development of knee OA and lateral chondral injury showed no relationship. There were ten other studies,23 ,25 ,33 ,35 ,37 ,49 ,50 ,60 ,66 ,73 of which one low risk of bias study35 showed conflicting evidence if the location of the chondral injury was not reported.
In nine studies,20 ,35 ,49 ,55–57 ,67 ,68 ,75 of which two were low risk of bias studies, a distinction between medial and lateral meniscus injury/meniscectomy was made. We found moderate evidence for a positive relationship between medial meniscus injury/meniscectomy and development of OA (tibiofemoral and unspecified) in patients with an ACL rupture. Conflicting evidence was found for influence of lateral meniscus injury/meniscectomy on tibiofemoral OA development and moderate evidence for no significant relationship on OA development in compartment unspecified. Twenty-one high risk of bias studies did not report the location of the meniscus injury; these studies showed limited evidence for positive relationship with development of tibiofemoral OA and conflicting evidence if the compartment of OA development was unspecified. The studies did not report the extent of meniscectomy. Results of meniscus injury/meniscectomy showed conflicting evidence for a relationship with patellofemoral OA development. One low risk of bias study56 and one high risk37 of bias study reported no significant relationship, and in one high risk of bias study41 meniscus injury/meniscectomy was related to patellofemoral OA development.
In seven studies,37 ,42 ,43 ,56 ,60 ,61 ,66 one of them low risk of bias study, moderate evidence for no relationship was found for the influence of time between injury and reconstruction on development of tibiofemoral and patellofemoral OA. Seven studies did not specify the compartment of OA outcome and these studies showed conflicting evidence.25 ,35 ,36 ,68 ,74 ,78 ,79
In 13 studies investigating ACL reconstruction versus non-operative treatment, conflicting evidence was found with patellofemoral OA,19 ,31 ,41 ,47 ,70 tibiofemoral OA19 ,31 ,40 ,53 ,70 ,80 and if no specific compartment30 ,65 ,72 ,77 ,79 was reported.
Fourteen studies reported outcomes on the relationship between bone-patellar tendon-bone graft versus hamstring tendon graft and development of tibiofemoral OA or OA in compartment unspecified. The studies gave conflicting findings. Mascarenhas et al52 and Leys et al38 reported opposite results for the development of medial and lateral tibiofemoral OA; Mascarenhas et al found a positive relationship between bone-patellar tendon-bone graft and development of lateral tibiofemoral OA, whereas Leys et al found a positive relationship between bone-patellar tendon-bone graft and development of medial tibiofemoral OA. In six studies,18 ,19 ,37 ,38 ,52 ,69 the influence of graft type on patellofemoral OA was studied: limited evidence was found for no relationship. Conflicting evidence in two high risk of bias studies26 ,68 was found for the influence of allograft on OA development in compartment unspecified.
One low risk of bias study and five high risk of bias studies reported on the influence of tunnel placement of the ACL reconstruction and OA development. Two studies showed no significant relationship between tunnel placement and patellofemoral OA development.56 ,62 Four high risk of bias studies38 ,46 ,58 ,68 evaluated the influence on development of OA in compartment unspecified; three studies38 ,46 ,58 found no significant relationship, resulting in limited evidence for no relationship.
Two studies with high risk of bias reported on the influence of double-bundle and single-bundle ACL reconstruction and OA development in compartment unspecified.25 ,27 These studies showed limited evidence for no relationship with development of OA.
We summarised the available evidence concerning which determinants influence the risk of OA after ACL rupture. Sixty-four studies were included, but 62 were classified as high risk of bias.
Key clinically relevant findings
There was moderate evidence for:
Medial meniscal injury/meniscectomy influencing OA development (tibiofemoral OA and compartment unspecified).
No relationship with time between injury and reconstruction and OA development in patellofemoral and tibiofemoral compartments.
No relationship between OA development in unspecified compartments and the following determinants was found: sex, laxity, range of motion and lateral meniscal injury/meniscectomy.
There was limited evidence for influencing OA development by the following determinants:
Medial and patellar chondral injury (compartment unspecified).
Meniscal injury/meniscectomy if the location was not reported (tibiofemoral OA).
Meniscectomy of both menisci (compartment unspecified).
Meniscectomy in non-operatively treated patients.
The following determinants showed limited evidence for no relationship with OA development:
BMI (patellofemoral OA).
Graft type (patellofemoral OA).
Activity level pre-reconstruction (compartment unspecified).
Lateral chondral injury (compartment unspecified).
Tunnel placement (patellofemoral OA and compartment unspecified).
Single-bundle versus double-bundle ACL reconstruction technique (compartment unspecified).
Notably, most studies reported only radiological OA. Only two studies43 ,70 reported both radiological OA and clinical OA as outcomes for evaluating the influence of determinants. Thus, the findings of this systematic review address the influence on radiological OA but not on clinical OA. We were also interested in determinants that influence early degenerative changes; however, the majority of the included studies reported mid-term or long-term follow-up. A mean follow-up time ≤5 years was reported in only eight studies.
The role of the meniscus: keep or cut?
Many studies evaluated the influence of the meniscus on the development of OA. The majority of studies did not report the location of the tear, the extent of meniscectomy, and in which compartment the OA was developing. We had no information about the influence of the time of the meniscus injury, also a possible confounder.
Although more extended, our results are in line with the findings of the previous reviews concerning meniscal injury and meniscectomy as risk factors for tibiofemoral OA development. However, these previous reviews did not distinguish between medial and lateral meniscal injuries/meniscectomies.
Our review provides important data that medial meniscal injury/meniscectomy showed a relationship with the development of OA, but lateral meniscal injury/meniscectomy did not. Anatomically, the medial meniscus is more rigid with less anterior posterior mobility than the more mobile lateral meniscus; this could have an effect on the secondary OA changes of the affected compartment.82
These findings contradict the results of a systematic review concerning clinical outcome and risk of OA development in patients undergoing meniscectomy. In that review, Salata et al83 found four studies with a higher rate of OA in the lateral meniscectomy group, two studies reporting no significant difference, and one study in which medial meniscectomy was more related with OA. These results were not included in our systematic review because the meniscus studies did not meet the inclusion criteria. Moreover, most studies did not report the location (medial or lateral compartment) of the meniscal resection, making it difficult to discern the specific influence of medial/lateral meniscectomy.
A possible explanation for conflicting evidence for development of OA (compartment unspecified) and limited evidence for a positive relationship with development of tibiofemoral OA is the heterogeneity of the location of meniscectomy. Also, the included studies did not report the extent of meniscectomy, except the study of Fink et al,30 which found in patients treated non-operatively for their ACL rupture a significant correlation between the degree of OA and the amount of meniscal resection that was performed at the time of the initial arthroscopy. For the ACL reconstructed group, there was no significant correlation.
A focus on patellofemoral OA
Patellofemoral OA is gaining consideration as an important clinical entity.84 Regarding OA of the patellofemoral joint, two studies37 ,56 found no relationship with meniscal injury/meniscectomy in an ACL reconstructed population. However, in the study of Keays et al,37 the relationship was close to significant and in another study meniscal injury/meniscectomy was significantly associated with patellofemoral OA.41 Furthermore, in a population without ACL injury, meniscectomy was related to development of patellofemoral OA.85 An explanation for this relationship could be the influence of altered biomechanics in the knee, or the meniscal tear was a feature of the already existing early knee OA.
The results of this systematic review confirm the thoughts about the importance of preservation of the meniscus for preventing development of OA. Our advice for future studies is to document the location and extent of meniscectomy as well as which knee compartments, medial, lateral or patellofemoral, were used for assessing OA development.
Three key clinical questions and our findings
In clinical practice, three questions are important with regard to choice of treatment for ACL injuries and the development of knee OA.
What is the influence of operative versus non-operative treatment on OA development? On the basis of our results, we cannot answer this question because there was conflicting evidence. However, we should note that, in the operatively treated patients, the graft type was mostly bone-patellar tendon-bone.30 ,31 ,40 ,41 ,47 ,53 ,65 ,70 ,72 ,77 ,79 ,80 So, there is less information on hamstring tendon reconstructed patients versus non-operatively treated patients and development of OA, despite both graft types being commonly used for ACL reconstruction.86
When operative treatment is chosen, what is the influence of graft choice? On the basis of this systematic review, we cannot recommend one graft type to reduce OA risk.
Is early reconstruction necessary for preventing OA development? The aim of early timing of reconstruction after ACL rupture is to prevent new meniscal and cartilage damage. Our results indicate that early or late reconstruction is not related to greater risk of patellofemoral or tibiofemoral OA.
However, for OA development in unspecified compartment, we cannot give any indication which time point, early or late after injury, is best for reconstruction with regard to preventing OA development. A possible explanation for these conflicting results is the heterogeneity of additional injuries in the included studies and differences in the definition of early reconstruction. Furthermore, Smith et al87 found in their meta-analysis no significant difference in the incidence of chondral and meniscal injuries between early and delayed reconstruction groups (the latter was defined as a minimum of 6 weeks postinjury). Another explanation might be that degenerative changes develop after the initial trauma caused by, for example, traumatic bone marrow lesions and activation of proinflammatory cytokines, independently of the choice of treatment.3 Besides, ACL reconstruction is a new trauma with additional damage such as bone marrow lesions, haemarthrosis and inflammation-related factors, for example, inflammatory cytokines.
We did not distinguish between partial and complete ACL tears. Partial or complete tears need to be diagnosed by arthroscopic evaluation, the reference for diagnosing ACL rupture. We may assume that the studies that included operatively treated patients enrolled patients with complete ACL tears. However, most studies did not describe their arthroscopic findings. Of the four studies which included non-operatively treated patients, one64 reported the inclusion of both partial and complete tears, two22 ,76 reported the inclusion of only complete tears and one 71 did not describe the type of the ACL tear. Thus, it is difficult to draw conclusions about the difference between the influence of partial and complete tears on OA development. Besides, in long-term follow-up studies, it is possible that partial tears progress to complete ACL tears88 and then it is difficult to distinguish the contribution of the partial and complete tears to the development of OA.
A determinant, which was not included in the results, is the altered knee biomechanics after ACL injury. The possible explanation for no information about this determinant is that studies researching the altered knee biomechanics include fewer patients (n≤20, exclusion criteria of this systematic review) and that these studies have a cross-sectional design (exclusion criteria of this systematic review). Chaudari et al89 suggest that the observed changes in the knee biomechanics result in altered loading patterns and influence metabolic changes in the underlying cartilage. Reduced internal tibial rotation was found in patients after ACL reconstruction compared with the contralateral knee and healthy controls.90 In addition to this finding, a recently published cross-sectional study showed that after ACL reconstruction, patients with patellofemoral OA and valgus alignment had significantly less internal knee rotation during walking and running than patients with valgus alignment and no patellofemoral OA.91 However, this study had a cross-sectional design; prospective studies are required to evaluate if the altered knee rotation is a result of patellofemoral OA or influences the development of patellofemoral OA.
This systematic review has some limitations. First, of the 64 included studies, only 1423 ,26 ,34 ,35 ,38 ,42 ,43 ,46 ,47 ,56 ,65 ,67 ,68 ,70 corrected for the influence of confounders. Consequently, the reported influence of determinants on the development of OA may be partly or completely explained by other factors. By presenting the data, one of the criteria to be classified as a low risk of bias study was controlling for confounding. A prospective observational study design is the best way to determine predictors for development of OA after an ACL rupture. However, prospective collected data and retrospective analyses (research question defined after data collection) were also useful for our research question. Therefore, we also included retrospective study designs.
Second, the number of patients available for analysis at follow-up in the included studies was small. Only 18 of the 64 (28%) included studies had more than 100 patients available for analysis at follow-up.
Third, the included studies were heterogeneous with regard to study design, determinant assessment, additional intra-articular injuries, reported OA outcome, definition of OA and the statistical methods used. For these reasons, comparison between the included studies was difficult and pooling of the data was not possible. Therefore, we used the second best option for presenting the results: best-evidence synthesis.
Best-evidence synthesis is appropriate for summarising the available evidence. All the 64 included studies were classified as low risk or high risk of bias; however, only two studies met the criteria for low risk of bias. This means that reporting of inclusion of consecutive patients, measurement of determinants and outcomes independently, using accurate measures for the determinants and description of loss to follow-up with a maximal 20% and correction for confounding were poorly performed and described in the included studies.
Finally, we attempted to evaluate the influence of determinants on the development of tibiofemoral and patellofemoral OA separately. However, we should note that some studies did not use a valid tool for the compartmental assessment of OA, (eg, Kellgren and Lawrence score for assessment of patellofemoral OA). In some studies, the compartment was not described (compartment unspecified). The evaluation of the correctly used classification system for compartmental OA assessment was not included in the quality assessment tool.
The strengths of this systematic review are that we summarised the evidence for tibiofemoral OA and patellofemoral OA outcomes after ACL injury separately. Moreover, we summarised these outcomes in patients who had had ACL reconstruction and those who had been managed with conservative treatment. Additionally, we evaluated determinants that influence early degenerative changes because we included studies with relatively short follow-ups (a minimum of 2 years). To be comprehensive, we chose to include both prospective and retrospective study designs having at least 20 patients. In addition to previously published systematic reviews,5–7 we included 21 studies published after the search dates of those systematic reviews.
Studies that used outdated surgery techniques were excluded, which resulted in exclusion of many older studies. However, our oldest study included was published in 198964 and newer studies might be of better quality as our two low risk of bias studies were published in 201256 and 2013.35 The best-evidence synthesis considers the quality of the studies and accounts for a possible bias. When we analysed the results of studies only published during the past 10 years, the results differed minimally. The only aspects that changed were the influence of chondral injury (location not reported) on OA development (compartment unspecified), and of the graft type bone-patellar tendon-bone; both would change from conflicting evidence to limited evidence for a positive relationship with development of OA. These results of limited evidence still need more high-quality studies in order to make firm recommendations.
Overall, we can conclude that despite the inclusion of many new studies in this comprehensive systematic review, including two low risk of bias studies,35 ,56 more low risk of bias studies are required to evaluate determinants and their role in OA development. Many determinants showed conflicting and limited evidence. The following determinants should be further studied in large prospective studies, which could be used for meta-analysis: knee function and activity level, both examined in the first period after ACL rupture, patients characteristics, such as age, BMI and sex, meniscal injury/meniscectomy specified in medial and lateral compartments, meniscus repair, chondral injury, choice of treatment, graft type and reconstruction technique. We strongly recommend specifying the compartment of OA development.
In summary, medial meniscal injury/meniscectomy after ACL rupture influences the development of OA (tibiofemoral OA and compartment unspecified). In contrast, it seems that lateral meniscal injury/meniscectomy has no relationship with OA development. Our results also suggest that time between injury and reconstruction does not influence the development of patellofemoral and tibiofemoral OA. However, we found limited or conflicting evidence for many determinants.
What are the new findings?
In patients with an anterior cruciate ligament rupture:
Moderate evidence was found that medial meniscus injury/meniscectomy had influence on osteoarthritis (OA) development; in contrast, lateral meniscus injury/meniscectomy showed moderate evidence for no relationship with development of OA.
Time between injury and reconstruction showed moderate evidence for no relationship with patellofemoral and tibiofemoral OA development.
It is still unclear which treatment option is the best for preventing OA development; conflicting evidence was found between treatment choice (operative vs non-operative treatment) and development of knee OA.
The authors would like to thank Louis Volkers for his help with the literature search.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online table 1
- Data supplement 2 - Online table 2
- Data supplement 3 - Online table 3
Contributors All authors were involved in designing this systematic review, interpreting the data and have contributed to the writing and editing of the manuscript. BLvM and MR assessed the studies for inclusion. BLvM and WAvE performed the data extraction. MR, SMAB-Z, DEM, JANV and BLvM performed the risk of bias assessment.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.