Risk of knee replacement after anterior cruciate ligament injury: a systematic review

BACKGROUND: It is recognized that patients suffering rupture of the anterior cruciate ligament (ACL) are prone to accelerated osteoarthritis (OA), thus, this patient group is at increased risk of undergoing knee replacement (KR) at a younger age compared to the general population. There are limited data directly determining the risk of KR in ACL-injured patients. The purpose of this study is to estimate the risk of undergoing KR in patients with a history of primary ACL injury (ACLi). STUDY DESIGN: Systematic review. METHODS: A systematic literature search with keywords and MeSH terms was performed using PubMed and Cochrane Library. We identied all studies reporting KR as an outcome in individuals with primary ACLi. Individuals with concomitant and secondary intra-articular injuries were included. A modied version of the Coleman Methodology Score (CMS) was used to evaluate the quality of the included studies. The risk was estimated by using the total number of ACL-injured patients and the total number of KRs within this cohort, consisting of both patients with reconstruction and patients with non-operative treatment. RESULTS: A total of 152 390 ACL-injured patients and a total of 1071 KRs were assembled from 5 prospective and 7 retrospective studies, which revealed a mean modied CMS of 70. Mean follow-up time was 16 years (ranging from 4.6 to 30 years). Mean age among the patients was 29.8 years. Estimated risk of undergoing KR after ACLi was 0.7 % with values between 0.54 % and 9.5 % among the included studies. CONCLUSION: This systematic review suggests an estimated risk of KR among ACL-injured patients to be 0.7 %. Our ndings are based on the results from the included studies, which, to some degree, showed divergence. The discrepancy may be due to variability in several factors, both research variables and patient factors, including follow-up time, sample size, subsequent injuries, obesity, sex, age and activity level. No consensus about OA development after ACLi has been established, however, OA is a common topic in the conversation between the doctor and the patient. This data may be helpful for physicians and clinicians when informing patients about the long-term

the aim of enabling return to sports by restoring knee stability and function (23)(24)(25). Filbay et al. (2015) identi ed no signi cant difference in knee-related QoL between reconstructed patients and conservatively managed patients (16).
On the contrary, a recent study by Sanders et al. (2016) comparing reconstructed patients and patients treated non-operatively with a mean follow-up time of 13.7 years, found that ACLR signi cantly reduced the risk of secondary MT and OA compared with non-surgical treatment. The authors of this study criticize the short follow-up time in the study by Frobell et al. (2010), as the results of their analysis demonstrate that there is little difference in arthritis between patients treated with ACLR and nonoperatively treated patients until approximately 6 years after ACLi. Additionally, non-operatively treated patients were 5 times more likely to sustain secondary MT than reconstructed patients (26). A comment on the study by Frobell et al. (2010) from Løken et al. (2011) also addresses the 5-year follow-up as a weakness, bringing up the subject of long-term results (27). In the unstable knee, reconstruction may prevent subsequent injuries by reestablishing su cient functional stability (28), indirectly protecting against accelerated knee OA, but reconstruction is listed as an arthrogenic factor, as it is possible that the procedure can kick-start the development of OA (11). A systematic review by Sommerfeldt et al. (2018) found a high incidence of knee OA after ACLi, which partially may be explained by the fact that an ACLde cient knee is more prone to subsequent damage to cartilage and menisci (29). Moreover, a recent study by Lindanger et al. (2019) concludes with an increased risk of KR among patients with late ACLR (≥ 24 months) compared to the group with early ACLR (30). There is a lack of evidence regarding the superior treatment due to the discrepancy in available literature. Treatment of ACLi remains controversial in 2020.

Knee osteoarthritis
OA is a common degenerative disorder characterized by loss of cartilage in the joints with past trauma as a risk factor. Other risk factors include advancing age, obesity, genetics and female sex. Age is the strongest predictor of radiographic OA development and progression. The prevalence of joint pain is markedly increased with age in the general population. OA is primarily a clinical diagnosis; the association between radiographic features of OA and joint pain is not constant (31-33).

Traumatic knee injuries accelerate knee osteoarthritis development
A key concern after knee injuries is the development of OA. Knee OA is a signi cant cause of impaired mobility (33). Progression of knee OA is slow, generally taking several years. Acute joint injuries, such as ACLi and MT, provide a greater risk of early development of knee OA. It is widely accepted that joint instability following ACLi causes meniscal lesions, and that meniscectomy, which results in loss of the load-bearing and shock-absorbing functions of the menisci, is associated with accelerated OA development (34)(35)(36)(37)(38). Long-term changes, such as disruption of normal mechanics in the joint and altered load distributions, come in addition to the intra-articular pathogenic processes initiated at the time of injury, contributing to the increased incidence of later OA (32). Furthermore, Khan et al. (2019) did the rst epidemiological study in quantifying the important link between ACLi and MT and end-stage OA treated by total knee replacement (TKR). They found an estimated 7-fold and 15-fold increased odds for TKR after ACLi and MT, respectively, when compared to an age-matched cohort, which demonstrates that ACLi and MT are signi cant independent risk factors for end-stage OA requiring TKR. Patients undergoing TKR after ACLi or MT were found to be signi cantly younger at the time of surgery than individuals without these injuries (39).
Additional individual variables, including genetics, sex, age, muscle strength, obesity, activity level and reinjury, contribute to variation in outcome. As a majority of individuals with acute ACLi are younger than 30 years at onset of injury, and several of them under 20 years, ACLi is responsible for a high quantity of patients with early-onset OA, contributing to associated pain, functional limitations and reduced QoL in the patient group between 30 and 50 years of age; the young patient with the old knee (6).  states that ACL rupture, isolated or with concomitant injuries to the menisci or collateral ligaments, results in radiologic changes indicating OA in 60-90 % of these patients 10-15 years after injury (40)(41)(42)(43)(44)(45), however, comparisons are uncertain due to different criteria for radiologic OA, as well as the small numbers of patients in the respective studies (46). Moreover, a later study by Lohmander et al. (2007), suggests that, on average, 50 % (ranging from 10-90 %) of those with a diagnosed ACLi or MT have OA with associated pain and functional impairment 10-20 years after the diagnosis (6).
In a review of risk factors of OA in ACL-injured patients, the authors list possible factors contributing to the origin of OA in patients with ACLi: (1) concomitant damage to menisci, articular cartilage and ligaments, (2) subsequent (traumatic) injuries due to instability and "giving way", which have an impact on the menisci and cartilage, as well as possibly stretching peripheral ligamentous structures and (3) chronic injury as a result of alterations in the biomechanics due to shearing forces affecting the articular cartilage (11).

Radiographic diagnosis
No universal methodological radiologic classi cation method exists (21), but the Kellgren-Lawrence system for radiographic diagnosis of OA has been the standard for many decades and is currently the most widely used clinical tool for radiographic grading of OA. Radiographic features, including osteophytosis, joint space narrowing (JSN), joint line sclerosis and subchondral cysts, are used to grade the OA severity from 0 (normal joint) to 4 (complete joint space loss) (47,48). Symptoms, Function in daily living (ADL), Function in Sport and Recreation (Sport/Rec) and knee-related QoL (50).

High physical activity as a risk factor
There is an association between participating in elite sports and an increased risk of OA. The degree of risk is dependent on the nature of the sport. Sports that involve high intensity and repetitive and high impact forces through the affected joints provide the greatest risk (51 (53). Similarly, results from a cohort of female soccer players between 26 and 40 years of age sustaining ACLi 12 years earlier, showed that more than 50 % of the participants had radiographic OA in their injured knee (54).
Even after ACLR, there is a 30-40 times greater risk of re-rupture in a young patient returning to sport, compared with an uninjured counterpart (55). According to a report from NKLR, hazard ratios for revision and contralateral ACLR for the [15][16][17][18][19] year age group are 4.0 and 4.9, respectively, compared with the patient group over 30 years of age, making young patients sustaining ACLi at greater risk of earlier post-traumatic osteoarthritis (PTOA) due to additional OA risk factors, which in this case are re-rupture and contralateral ACLi (56).

Prevalence of knee OA in Norway
From a population survey in Norway, looking at hand, hip and knee OA, the overall prevalence of knee OA was 7.1 % (57). A strength of this study is the limited geographical area, as all subjects were inhabitants of the Ullensaker municipality, making the population relatively homogenous. However, there are geographical differences, as the prevalence of OA among men is higher in the countryside, possibly explained by more heavy manual work in these areas (58).

Knee replacement
Knee replacement (KR), including unicompartmental knee replacement (UKR) and TKR, is an effective surgical intervention for end-stage OA, and is commonly performed (59). TKR has become the gold standard for management of disability and persistent pain associated with knee OA (60). KR is an increasingly common procedure, possibly due to an increasingly ageing society, as well as rising rates of obesity, causing an inevitable increase in rates of knee OA, which is the main clinical indication for TKR. One study suggest the lifetime risk of TKR for individuals over 50 years of age to be 8.1 % for men and as high as 10.8 % for women (61). Generally, patients undergoing KR have end-stage OA, but the decision to go under the knife should be guided by the patient's symptoms, although there is no consensus on the severity level of symptoms regarding surgery indication (62,63). The radiologic rate of OA after ACLi is noticeably higher than the symptomatic rate, still the latter is the most crucial to the patient (11). Overall, KR is a powerful and evident surrogate endpoint for severe OA (64,65). 7000 KR procedures are annually performed in Norway. The number is increasing, and it has been doubled from 2005-2018. The majority of patients have OA with a mean age of 69 years for women and 67 years for men when undergoing KR. 63 % of the patients are women. In 2018, 223 KR procedures were due to sequelae after ligament injury (66).
A study by Leroux et al. (2014) comparing cumulative incidence of KR between patients with ACLR and a non-injured control group, found the cumulative incidence of KR following cruciate ligament reconstruction after 15 years to be low (1.4 %), however, it was 7 times greater than the matched control subjects from the general population (0.2 %) (59). There are few existing studies analyzing KR as an outcome in ACL-injured patients. The purpose of this systematic review was to estimate the risk of undergoing KR among patients with a history of primary ACLi with or without subsequent lesions.

Search strategies
Systematic research started January 2019 in PubMed and Cochrane Library. Relevant MeSH terms for the topics "anterior cruciate ligament injury" and "knee replacement" were looked up using PubMed's MeSH database. Different search techniques were used to narrow the results and nd the most relevant articles: truncation was added to the word stem to nd the plural variations of a word, i.e. injur* includes both injury and injuries, quotation marks were used to keep terms together and in order, i.e. "knee replacement" vs. knee replacement, parentheses for combining concepts and the Boolean operators AND and OR were used to connect and de ne the relationship between the search terms.
Using the advanced search page in PubMed, each topic was searched separately before combining them together in one search: ("anterior cruciate ligament injur*" OR "anterior cruciate ligament rupture" OR "anterior cruciate ligament tear" OR "ACL injur*" OR "ACL rupture" OR "ACL tear") AND ("knee replacement" OR "joint replacement" OR arthroplast*). The same search strategy was performed in Cochrane Library.
The number of studies included in this thesis increased after updating the search December 2019, as well as after reviewing the reference lists of the relevant studies.

Study selection
We included any study reporting KR, both UKR and TKR, in patients with history of primary isolated ACLi; however, concomitant collateral ligament sprains and meniscal injuries were also included, as this will be a part of the primary injury resembling the impact of soft tissue injuries on the outcome of a fracture. Due to no consensus regarding the long-term protective role of ACLR against knee OA, we decided to merge patients with ACLR and patients without ACLR, which represented one case group in our study.
Since OA is an age-related disease, studies reporting the cohort with mean age over 50 years were excluded. Other exclusion criteria were age under 15 years when sustaining ACLi, previous knee injury and previous knee surgery. Having a previous knee injury/surgery would have an impact on the outcome that would affect the interpretation regarding the role of ACLi; the more injuries to the joint, the greater the risk of accelerated knee OA. We wanted to see the relation between rupture of ACL and KR, hence, we included only subjects with primary ACLi with or without subsequent injuries. As already commented on, subjects with secondary damage, such as MT and re-ruptures, were not excluded, as it is recognized that ACLinjured patients are at an increased risk, and furthermore, to capture the infrequent event of KR.  (52). We decided to exclude this study in our review, as the rate of OA progression following knee injury is accelerated in this otherwise healthy, high-demand patient population (11). Thus, including a study examining a patient group consisting exclusively of high level athletes would have given a too high risk when determining the risk of KR in patients with activity level at an average level, although we are aware of the fact that a large number of ACLi happen during sports activities and that an estimated 55 % return to their pre-injury activity level (14).

Data extraction
To estimate the risk of KR among patients with a history of ACLi, the number of ACL-injured patients and the number of KRs within this patient group extracted from each included study, were incorporated to get a total number of ACL-injured patients and a total number of KRs. Mean follow-up time, mean age and percentage of male within the cohorts were reported. One study (59) did not state mean age; the author of this systematic review manually calculated the mean age using the information provided.

Quality assessment
In evaluating the quality of the included studies, we used the methodology score introduced by Coleman et al. (77), which was later modi ed by Jakobsen et al. (78) (Appendix 1). The Coleman Methodology Score (CMS) originally consisted of 10 criteria and was developed to assess clinical studies on surgical treatment of patellar and achilles tendinopathy with a total score ranging from 0-100 (21,79). Achieving a score of 100 indicates that the study contains minimally confounders and other biases. The difference between the original CMS and the Modi ed Coleman Methodology Score (MCMS), is that MCMS describes the post-operative rehabilitation, while CMS only evaluates compliance of the rehabilitation protocol among the participants (80).
Since no cutoff for MCMS to de ne high or low methodological quality has been reported (21), we decided, after assessing the quality of the eligible studies, giving us a subjective impression of acceptable quality, that a minimum score of 50 was required to be included. The literature suggests a score of more than 55 % for other checklists to be considered as high quality (81), thus, we consider a minimum score of 50 to be reasonable. We are aware that the scoring system may be subject to different interpretations, as some of the criteria was not clearly de ned, thus, the scoring is prone to subjectivity and the score outcome is dependent on the investigator.

Statistical analysis
By combining the number of patients with ACLi from all of the eligible studies and the number of KRs within this patient group, we estimated the risk of undergoing KR after suffering an ACLi. After reviewing the included individual studies, a total of 152 390 ACL ruptures and 1071 KRs were included in our calculation. As mentioned previously, our cohort consisted of both patients with ACLR and patients with non-operative treatment; we estimated a combined risk of KR for these two patient groups. In addition, 95 % con dence interval (CI) was calculated for each study 3 Results Using the search strategies in PubMed and Cochrane Library, as well as reviewing the reference lists in relevant studies, abstracts of a total of 356 studies were reviewed, resulting eventually in 12 included studies. The study included from a reference list (59) did not appear in the systematic searches because of the missing citation terms "anterior cruciate ligament injur*/tear/rupture" or "ACL injur*/tear/rupture". Studies reporting KR as an outcome among patients with a history of primary ACLi, with or without ACLR, were included. Figure 2 shows the ow diagram illustrating the study selection. The characteristics of the included studies are presented in table 5. 3.1 Methodological quality Table 4 presents the results of the quality assessment of the studies using MCMS. Overall, the quality of the included studies was of high-quality with a mean score of 70 (ranging from . None of the studies achieved the maximum score of 100; however, two studies (68, 69) revealed a MCMS of 92.

Risk of undergoing KR
The total number of participants was 152 390, with sample sizes ranging from 41 to 111 212 patients. The total number of KRs was 1071. Risks ranged from 0.54 % to 9.2 % in the included studies. The risk of undergoing KR among ACL-injured patients was calculated by using the number of subjects and the number of KRs from each study to get a nal sum, giving a risk of 0.7 % (1071/152 390).

Demographic data
Mean follow-up time, percentage of male included and mean age at onset of study for each included study are presented in table 5. The study that did not report mean age (59), reported the median age to be 28 years, however, the mean age was manually calculated using the provided information about the percentage distribution among the age groups.

Conclusion And Discussion
The absolute rate of KR among ACL-injured patients varies in the literature. Though there is a reason to believe that this knee injury increases the general risk of OA in the individual, which also is re ected in the disability score from the government from April 1997 (19). The variations may be due to variability in follow-up time and sample size in the studies, but also covariates including secondary injuries, interventions, rehabilitation protocol, age, sex, as well as modi able factors such as activity level, BMI etc., which must be taken into consideration when interpreting the results. For instance, Hagmeijer et al.
(2019) had the highest reported risk of KR, which most likely is explained by additional meniscal injury as inclusion criterion in their study, along with a small sample size. Study population size is a crucial consideration; research including larger sample size has a higher signi cance level of the ndings (82). The standard error is determined by the sample size: the greater the selection, the less errors (83). Thus, larger sample sizes will provide more accurate values and reliable results, which are more likely to be representative of the population (84,85). Sample size is of great importance when considering the signi cance of the ndings, but another important factor when estimating the risk is follow-up time, as it takes several years to develop OA. Older age is the greatest risk factor for OA, but the disease severity is more often than not related to other risk factors of OA, e.g. joint injury, anatomical factors affecting joint mechanics, genetics and obesity (86). OA development in younger people is often due to an underlying reason, most likely as a direct result of a trauma (87). Thus, patients suffering ACLi have a higher risk of undergoing KR earlier, as well as an elevated lifetime risk of KR, due to accelerated OA development. The limited follow-up period in the included studies makes it di cult to determine the overall KR risk among patients with ACLi. It is highly possible that a longer follow-up time would have given an increased risk estimate. Nonetheless, it is always a certain risk of arthrosis in every human being; OA is an age-related disease and is generally present, to some degree, in individuals exceeding 60 years of age (48).
In this systematic review we found the estimated risk of undergoing KR among patients with a history of primary ACLi to be 0.7 %, which is consistent with the studies analyzing the largest samples (59,72). The included studies are not comparable due to the great variability of several factors, both baseline characteristics and study variables, thus, stating a risk of KR is di cult. This study has several limitations. As mentioned previously, the estimated risk is applied for both patients with reconstruction and patients with non-operative treatment, thus, differences in graft types used in the group with ACLR were not taken into account. In 2020, it is our understanding that the protective role of ACLR is still uncertain. Additionally, we did not report the number of secondary injuries, nor whether meniscal repair or meniscectomy was performed, despite the fact that meniscectomy is proven to represent a signi cant risk factor for knee OA (89,90), unlike meniscal repair, which is of great importance in preventing OA development in the knee joint (91); preserving as much functional meniscal tissue as possible in surgical treatment of meniscal lesions is desirable (92).
Another limitation of signi cance is the use of different radiographic classi cation systems in OA diagnosis, in either isolation or in combination with symptomatology, when the decision to undergo joint replacement is made. Indications for undergoing KR were not consistently reported in the included studies, and radiographic or patients' reports about their symptoms were not available for analysis.
Radiographic OA does not always re ect the severity of symptomatology; nonetheless, joint replacement does indicate severe symptomatic OA. Moreover, the included studies had different purposes (presented  in table 4). Only one study had KR as main outcome (59). Whether this is of signi cance for our ndings, is uncertain.
A trend of increasing incidence of ACLi and MT is observed (93,94), and furthermore, according to a recent study, the rate of pediatric ACLi has been increasing approximately 2.3 % annually for the past two decades (95). Individuals sustaining ACLi are predisposed to accelerated OA, and not surprisingly, injury at a young age very likely results in earlier OA development. Given this, and the fact that PTOA is a consequence of joint trauma, it is reasonable to assume that sequelae of these injuries constitute a signi cant amount of TKRs performed. Reducing the demand for arthroplasty is favorable, thus, the need for new strategies in OA prevention following knee injury, as well as management of early-stage OA, is emphasized. When ACLi has occurred, our work will be minimizing undesirable outcomes, such as joint replacement surgery, as well as informing patients about the risk. OA development after ACLi is multifactorial, but rehabilitation protocols, including neuromuscular knee exercise, should be initiated in every patient (96). Physicians and clinicians should be aware of modi able factors that could postpone the need for KR, hence, give advice about lifestyle changes. There is a lack of knowledge in the eld and patient preference of knowledge is often guided by the choice of top level athletes. This might not be the best role model for the general patient subjected to this knee injury. Further research, either from knee registries or long-term follow-up in randomized controlled trials, might be key factors to outline this warranted knowledge.
Something to consider is the trend of a higher threshold to perform joint replacement surgery in younger patients, which may underestimate the burden of disease (96). Although the absolute rate of KR following primary ACLi is relatively low, the risk of undergoing KR at a younger age is signi cantly increased, compared to the general population (72). Overall, available literature reporting KR among patients with primary ACLi is scarce, and the Coleman scorings reveal that there is room for improvement in research to outline the risk of KR after ACLi. Furthermore, another signi cant limitation in this study is omitting the length of time from ACLi to KR, which is due to the absence of this important factor in most studies. We are fully aware of the fact that a risk estimate without including the time gap between injury and KR is of less value and that a methodology showing Kaplan-Meier curve would be optimal. Nevertheless, ndings in this systematic review may be helpful when counselling patients about the risk of undergoing KR after ACLi, but it is desirable with more high-quality, preferably prospective, studies with minimal variability in Consent for publication Not applicable.
Availability of data and materials All data and materials used in this study are available online.

Competing interests
The authors declare that there is no con ict of interest.   (11). This is consistent with OA being a chronic disease progressing over time and that OA development is accelerated by high activity. Furthermore, this cohort study shows a protective role of ACLR, which may be explained by the reduced risk of secondary MT and/or chondral lesions when participating in elite sport

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.