Article Text

Download PDFPDF

Association between MRI-defined osteoarthritis, pain, function and strength 3–10 years following knee joint injury in youth sport
  1. Jackie L Whittaker1,2,3,
  2. Clodagh M Toomey3,
  3. Linda J Woodhouse1,
  4. Jacob L Jaremko4,
  5. Alberto Nettel-Aguirre3,5,6,7,
  6. Carolyn A Emery3,5,6,7
  1. 1 Department of Physical Therapy, Faculty of Rehabilitation Medicine, University of Alberta, Edmonton, Alberta, Canada
  2. 2 Glen Sather Sports Medicine Clinic, University of Alberta, Edmonton, Alberta, Canada
  3. 3 Sport Injury Prevention Research Centre, Faculty of Kinesiology, University of Calgary, Calgary, Alberta, Canada
  4. 4 Department of Radiology and Diagnostic Imaging, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada
  5. 5 The Alberta Children’s Hospital Research Institute for Child and Maternal Health, Calgary, Alberta, Canada
  6. 6 Department of Pediatrics, University of Calgary, Calgary, Alberta, Canada
  7. 7 Department of Community Health Sciences, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada
  1. Correspondence to Dr Jackie L Whittaker, Faculty of Rehabilitation Medicine, University of Alberta, Edmonton, Alberta T6G 2G4, Canada; jwhittak{at}


Background Youth and young adults who participate in sport have an increased risk of knee injury and subsequent osteoarthritis. Improved understanding of the relationship between structural and clinical outcomes postinjury could inform targeted osteoarthritis prevention interventions. This secondary analysis examines the association between MRI-defined osteoarthritis and self-reported and functional outcomes, 3–10 years following youth sport-related knee injury in comparison to healthy controls.

Methods Participants included a subsample (n=146) of the Alberta Youth Prevention of Early Osteoarthritis cohort: specifically, 73 individuals with 3–10years history of sport-related intra-articular knee injury and 73 age-matched, sex-matched and sport-matched controls with completed MRI studies. Outcomes included: MRI-defined osteoarthritis, radiographic osteoarthritis, Knee Injury and Osteoarthritis Outcome Score, Intermittent and Constant Osteoarthritis Pain, knee extensor/flexor strength, triple-hop and Y-balance test. Descriptive statistics and univariate logistic regression were used to compare those with and without MRI-defined osteoarthritis. Associations between MRI-defined osteoarthritis and each outcome were assessed using multivariable linear regression considering the influence of injury history, sex, body mass index and time since injury.

Results Participant median age was 23 years (range 15–27), and 63% were female. MRI-defined osteoarthritis varied by injury history, injury type and surgical history and was not isolated to participants with ACL and/or meniscal injuries. Those with a previous knee injury had 10-fold (95% CI 2.3 to 42.8) greater odds of MRI-defined osteoarthritis than uninjured participants. MRI-defined osteoarthritis was independently significantly associated with quality of life, but not symptoms, strength or function.

Summary MRI-detected structural changes 3– 10 years following youth sport-related knee injury may not dictate clinical symptomatology, strength or function but may influence quality of life.

  • adolescent
  • knee injuries
  • knee ACL
  • MRI
  • sport

Statistics from

Request Permissions

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.


Osteoarthritis (OA) is the fastest-growing global health condition based on years lived with disability.1 Although the aetiology of OA is multifactorial, previous joint injury is an established risk factor with meta-analyses indicating a 3.9-fold (95% CI 2.72 to 5.57) increased risk of radiographic OA 15 years following significant knee injury.2 Youth and young adults that participate in sport and recreational activities represent a subpopulation with increased vulnerability for knee injury and subsequent early-onset OA.3

For many years, OA diagnosis was based solely on structural radiographic findings (eg, joint space narrowing, osteophyte formation) classified by the Kellgren-Lawrence grading system (ie, grade ≥2).4 The clinical relevance of this diagnostic was questioned due to the discordance between radiographic features and clinical OA symptoms (eg, pain, reduced function)5 6 and the recognition that one does not seek treatment in the absence of pain or reduced function. This led to the evolution of a new diagnostic standard ‘symptomatic radiographic OA’ that requires some combination of pain and reduced function in addition to structural changes.7

Given the need to focus on primary (ie, prevent joint injury in susceptible populations) and secondary (ie, delay or halt OA onset after joint injury) prevention strategies to reduce the OA burden,8 researchers turned their attention towards an earlier structural OA diagnostic. It is now well documented that radiographic OA features are preceded by cartilage loss and bone marrow lesion (BML) formation, which are better detected with MRI.9 This has led to the development of various methods to measure structural MRI OA changes.10 One semiquantitative example of this is the MRI Osteoarthritis Knee Score (MOAKS).11 Recently, a definition of MRI-defined OA based on the MOAKS was proposed.12 However, the value of MRI-defined OA as an early OA diagnostic is unknown, and it is speculated that its use may lead to overdiagnosis based on incidental findings.6 11 13 14 Further, preliminary evidence would suggest that, like radiographic features, the relationship between the MRI OA features and clinical symptomatology and function is poor.15 Finally, there is a paucity of information about MRI-defined OA early in the period (3–10 years) following a youth knee joint injury.

The aims of this secondary analysis of a primary matched cohort study are to assess the relationship between MRI-defined OA and self-reported (ie, pain, symptoms, function, quality of life) strength and functional outcomes taking into consideration the influence of 3–10 years history of youth sport-related knee injury, body mass index (BMI), sex and time since injury.16 17



The Alberta Youth Prevention of Early Osteoarthritis (PrE-OA) longitudinal historical cohort study compares 100 youth/young adults (15–26 years) who sustained a youth sport-related intra-articular knee injury 3–10 years previously and 100 age-matched (within 12 months), sex-matched and sport-matched (at the time of injury) healthy controls on a variety of outcomes consistent with future OA. The analyses in this paper included a subsample of matched pairs for whom clinical MRI studies were available. Specifically, from the 168 participants (84%) that attended diagnostic imaging, 73 (n=146) matched pairs were identified.

The PrE-OA cohort sample size calculation, recruitment, injury diagnoses procedures and exclusion criteria have previously been reported.18 Briefly, injured participants had sustained an intra-articular knee injury during a previous cohort study or presented to a Sport Medicine Centre with a sport-related intra-articular knee injury 3–10 years previously when they were ≤18 years of age. Intra-articular knee injury was defined as a clinical diagnosis of knee ligament, meniscal or other intra-articular tibiofemoral or patellofemoral injury within the past 3–10 years that required both medical consultation and resulted in disruption of regular sport participation. Injury diagnoses were based on diagnostic codes recorded on injury report forms (physiotherapist clinical examination) or medical records (physician clinical examination) and confirmed by participants. Uninjured participants included individuals with no previous time-loss knee injury. Ethics approval was granted from the Conjoint Health Research Ethics Board, University of Calgary, Canada, and all participants provide signed consent/assent prior to testing.


Data were collected during one testing session. After completing a study questionnaire that gathered demographic, knee injury/surgery and medical history information, the Knee Injury and Osteoarthritis Outcomes Score (KOOS)19 and Intermittent and Constant Osteoarthritis Pain (ICOAP) score,20 individual participants rotated through testing stations that measured anthropometrics (height, weight), knee extensor/flexor isometric strength and dynamic balance. BMI (kg/m2) was calculated from height (0.1 cm; shoes removed) and weight (0.1 kg) assessed using a medical scale and stadiometer (Model 402 KL, Pelstar, McCook IN, USA). Background, measurement properties, testing procedures and data reduction for the KOOS, ICOAP, knee strength, triple-hop and Y-balance test (YBT) are reported in online supplementary appendix 1.

Supplementary file 1

All MRI and X-ray studies were conducted at an off-site diagnostic imaging facility. Participants underwent bilateral knee MRI using typical clinical sequences (ie, sagittal proton density, sagittal and coronal proton density fat saturated and 3D gradient echo FIESTA; 1.5 Tesla) and standard anterior–posterior (full extension and 30° flexion) and lateral (full extension) weight-bearing radiographs. Studies were rated by a musculoskeletal fellowship trained radiologist (JLJ) with 13 years of imaging experience, blinded to injury history or surgical intervention (other than those that required fixation that is visible on MRI). MOAKS scoring has very good to excellent reliability (kappa=0.61–1.0) across features in patients with significant OA features.11 Intrarater and inter-rater reliability of MOAKS scoring was confirmed by review of a subset of 40 MRI studies on two separate occasions (intrarater kappa (95% CI): BML 0.54 (0.19 to 0.89), cartilage 0.53 (0.21 to 0.86), osteophyte 0.73 (0.47 to 0.99), meniscus 0.72 (0.36 to 1.0), ligaments 0.47 (0.07 to 0.86)) and in comparison with another board-certified musculoskeletal fellowship-trained radiologist (inter-rater reliability kappa (95% CI): BML 0.56 (0.30 to 0.82), cartilage 0.44 (0.16 to 0.72), osteophytes 0.72 (0.36 to 1.0), meniscus 0.88 (0.64 to 1.0), ligaments 0.81 (0.55 to 1.0)). As synovitis effusion was infrequent in this sample, reliability of this subscale was not assessed. MRI-defined OA derived from MOAKS scores was based on established criteria (table 1).12 21 Specifically, participants were classified as having MRI-defined OA if they met the criteria for tibiofemoral (medial or lateral compartment), mixed tibiofemoral or patellofemoral MRI-defined OA. More information on the MRI sequences employed and MOAKS rating is contained in online supplementary appendix 1. Radiographic OA was defined as a Kellgren-Lawrence grade ≥2.4

Table 1

Criteria for MRI-defined OA derived from MOAKS ratings


Statistical analyses were performed using STATA V.12.1. In the case of missing data, the participant and their match were removed from the analyses. Covariates (ie, age, sex, BMI, sport, injury type and time since injury) are reported for both analysis groups (MRI-defined OA or not) as applicable (means or proportions (95% CI) or median (range)).

Mean (95% CI), proportion (95% CI) or median (range) and mean within-pair differences (95% CI) were calculated across outcomes as appropriate. ORs of MRI-defined OA by 3–10 years youth sport-related knee injury history (yes/no), injury type (patellofemoral subluxation/dislocation/grade I–III medial collateral ligament (MCL), lateral collateral ligament (LCL) or grade I–II ACL/isolated meniscal/grade III ACL and surgical reconstruction/grade III ACL and/or meniscal/bilateral) and surgery history (index knee surgery/bilateral knee surgery) were each assessed using logistic regression (OR, 95% CI) conditioned for the matched design. Multivariable linear regression models (clustered by matched pair) were used to assess the association between MRI-defined OA and each self-reported strength and functional outcome taking into consideration the influence of 3–10 years knee injury history (yes/no), BMI (overweight/obese or normal), sex (male/female) and time since injury (months). The value entered for the time since injury variable for control participants was the same as that of their matched case and reflected an equivalent injury-free time period. All assumptions of linear regression analyses were assessed and met. To ensure adequate adjustment for multiple comparisons, a significance level of α=0.004 was used (α=0.05/12 comparisons=0.004). A minimum clinically important difference (MCID) was considered as 6, 5–8.5, 7–8, 5.8–12 and 7–7.2 points for the KOOS pain, symptoms, function in daily living, function in sport and quality-of-life subscales, respectively,22 and 18.5 points for the ICOAP.23 MCID values for the triple-hop (per cent leg length), YBT (normalised to leg length) and knee extensor and flexor strength (Nm/kg) were not available.


The primary reason for not attending diagnostic imaging was accessibility and time restraints. Although there was a greater percentage of females who attended imaging (61%) than did not (55%), there was no difference in testing age (years: (imaging 22.3 (95% CI 21.9 to 22.7); no imaging 21.5 (20.9 to 22.2)), injury age (years: (imaging 15.5 (15.1 to 15.9); no imaging 15.0 (14.4 to 15.6)) or distribution of ACL and/or meniscal lesions (proportion: (imaging 38.0 (30.8 to 45.6); no imaging 31.3 (21.0 to 43.7)) between those that did and did not attend imaging.

Participant characteristics are summarised by MRI-defined OA status in table 2 and by injury history (ie, previous injury or no previous injury) in online supplementary appendix 2 (table 1). Participant median age was 23 years (range 15–27), and 62% were female. The median age of injury was 16 years (range 11–18), and the median time from injury to testing was 6.9 years (3–10). Thirty-eight per cent of knee injures were attributed to soccer, while the remaining were sustained in hockey, basketball, football, volleyball, rugby, running, baseball and downhill skiing. Of the 73 injured participants, 8 (11%) had patellofemoral subluxations, 16 (22%) had grade I–II MCL/LCL sprain or grade I–II ACL sprain, 10 (14%) had isolated meniscal injuries and 39 (53%) experienced grade III ACL sprains. All grade III ACL sprains were surgically reconstructed, and 63% presented with concomitant meniscal injuries.

Supplementary file 2

Table 2

Description of participant characteristics (n=146)

Twenty-four (16.4%) participants met the criteria for MRI-defined OA (21 previously injured and 3 controls), and six (4.1%) met the criteria for radiographic OA (all previously injured). Descriptive statistics across outcomes are summarised by MRI-defined OA in table 3 and by injury history in online supplementary appendix 2 (table 2). Due to missing data, two pairs were excluded for triple-hop (refused) analyses, one pair for YBT (refused) and ICOAP (not administered) analyses and three pairs for strength (dynamometer battery malfunction) analyses. There were no systematic differences (ie, age, sex, BMI or age of injury) between those that provided data and those missing data.

Table 3

Description of self-reported functional and strength outcomes by MRI-defined OA status (n=146)

The prevalence and ORs of MRI-defined OA by injury history, injury type and surgery history are summarised in table 4, while the distribution of MRI-defined OA is illustrated in figure 1. Injured participants were 10.0 times (95% CI 2.3 to 42.8) more likely to have MRI-defined OA compared with matched controls.

Table 4

Proportion and OR of MRI-defined OA by injury history, injury type and surgery history (n=146)

Figure 1

Compartmental distribution of MRI-defined OA. A total of 24 participants (3 controls and 21 previously injured participants) met the criteria for MRI-defined OA. *One of the participants that meet the criteria for patellofemoral MRI-defined OA also met the criteria for mixed tibiofemoral OA (ie, met the criteria for tibiofemoral OA only when features in the medial and lateral compartments were combined). In addition, 3 (12.5%) participants meet the criteria mixed tibiofemoral OA. OA, osteoarthritis.

The multivariable linear regression models examining the association between MRI-defined OA and each of the self-reported strength and functional outcomes are reported in table 5. The KOOS quality of life had a significant negative association with injury history and MRI-defined OA. Conversely, the KOOS pain, symptoms, function in daily living and function in sport/recreation subscales, ICOAP total or subscale (intermittent or constant pain) scores, normalised index knee extensor and flexor strength or triple-hop and YBT scores were not significantly associated with MRI-defined OA. Three to 10 years youth sport-related knee injury history was associated with poorer outcomes on the KOOS pain, symptoms, function in daily living and function in sport/recreation subscales, ICOAP total and intermittent subscale. Male sex was associated with greater extensor and flexor strength and triple-hop distance. Finally, BMI and time since injury were not significantly associated with any of the self-reported strength and functional outcomes.

Table 5

Associations between MRI-defined OA and self-reported function and strength based on multivariable linear regression


Three to 10 years history of youth sport-related intra-articular knee injury is associated with an increased OR of MRI-defined OA, which varies, by injury type and surgical history. Further, MRI-defined OA is significantly associated with self-reported knee-related quality of life but not self-reported knee pain, symptoms or function in daily living or sport and recreation, knee muscle strength or dynamic balance. These findings suggest that MRI-defined OA exists in young adults who have sustained a range of knee injuries (not exclusively ACL tears and meniscal lesions) and that these structural features may not dictate clinical symptomatology, knee muscle strength and functional performance but may influence knee-related quality of life 3–10 years following injury.

MRI-defined OA and knee injury

Many studies have identified an increased risk of knee OA after ACL tear, ACL reconstruction (ACLR) and/or meniscal injury.2 24 25 However, this conclusion was based primarily on investigations that have not considered healthy matched (age, sex and exposure) controls or individuals with other knee injuries. The current study supports the conclusion that those who sustain an ACL tear and/or meniscal lesion or undergo knee surgery are at a high risk of MRI-defined OA. However, it is important to highlight that MRI-defined OA is also present, although to a lesser extent, in individuals 3–10 years after a patellofemoral subluxation/dislocation, grade I–III MCL or LCL sprain and grade I–II ACL sprain, and as a result, these individuals should not be excluded from secondary OA prevention efforts.

Of interest, MRI-defined OA was present in three (4.1%) participants with no previous history of knee injury. This finding may be incidental due to the sensitivity of MRI for detecting preclinical lesions or it may be that these individuals exhibit structural changes due to aetiological factors other than previous injury (eg, genetic factors, nutritional deficits, undiagnosed disease processes, obesity, among others).26

MRI-defined OA and knee-related quality of life

Although previous investigations have identified quality-of-life deficits in persons that are ACL deficient,27 after ACLR,17 28 29 and in those with symptomatic radiographic OA,30 to the best of our knowledge the relationship between MRI-defined OA and knee-related quality of life 3–10 years postinjury has not been previously reported. Our analyses suggest that when considered in linear combination MRI-defined OA, injury history, BMI, sex and time since injury explain 40.3% of the variability in KOOS quality-of-life score. Specifically, that KOOS quality-of-life scores are statistically lower in participants with a previous injury (−7.5 points (95% CI −9.9 to –5.0), p<0.001) and those with MRI-defined OA (–6.5 points (−10.6 to –2.4), p=0.002) compared with those without an injury or MRI-defined OA. It is important to consider that the size of the observed changes is on the cusp of the MCID for the KOOS quality-of-life subscale (ie, 7–7.2 points) and therefore may not be clinically relevant.9 With that said, the MCIDs for KOOS subscales are unknown for youth and young adults; therefore, it is difficult to know if the observed differences are meaningful or not. Further, due to the limited number of persons with MRI-defined OA, the current analyses could not assess the interaction between injury history and MRI-defined OA; therefore, although both appear to impact knee-related quality of life independently, their additive effect 3–10 years following a youth sport-related knee injury remains unknown. Other factors that may contribute to knee-related quality of life include the ability to return to sport and subsequent and contralateral knee surgery.17 Accordingly, future investigations aimed at further elucidating the relationship between knee-related quality of life and MRI-defined OA following sport-related knee injury should aim for a larger sample size that enable multivariable analyses with more sophisticated models that consider the interaction of injury history and MRI-defined OA and other potential explanatory factors.

MRI-defined OA, pain, symptoms, strength and function

Although there is compelling evidence that articular cartilage injury and meniscal tear/treatment at the time of ACLR are predictive of KOOS scores 6 years post-ACLR,16 there is considerable evidence pointing to the discordance between structural changes (radiographic or MRI) and knee pain and/or clinical symptomatology.6 14 31 32 The lack of a linear association between MRI-defined OA and knee pain, symptomatology (eg, stiffness, function in daily living or sport and recreation), strength and function demonstrated in this investigation is consistent with these previous investigations and further questions the practice of using structural findings in isolation as a means to identify individuals with 3–10 years postknee injury that are likely to benefit from a secondary OA prevention intervention. With that said, it is plausible that the lack of evidence identifying a relationship between structural changes associated with MRI-defined OA and specific knee symptomology, knee muscle strength and dynamic balance tasks is because structural changes impact these outcomes differently across individuals and that the individual effect is lost when assessing for group differences. For example, in some individuals, it is plausible that MRI-defined OA is associated with pain, in others knee extensor weakness, and in others poor triple-hop performance, yet when these associations are combined across individuals they are lost. As knee-related quality of life has been shown to correlate with knee pain (r=0.60) and symptoms (r=0.79),18 physical functioning (r=0.29),33 social functioning (r=0.59) and general health (r=0.49),34 a plausible explanation for its association with MRI-defined OA is that, as it encompasses multiple constructs, it is more capable of demonstrating a group effect. A further consideration is that it is possible that only individual features of OA (ie, BML) are associated with poorer pain, strength and functional outcomes and these individual associations are lost when using a blended construct such as MRI-defined OA.

Implications for secondary OA prevention strategy screening

The findings of this investigation suggest that an early diagnostic for OA should consider the presence of structural features, intermittent clinical symptoms (eg, self-reported pain and stiffness) and functional deficits (eg, self-reported sport/recreation participation, strength or hop test). Similarly, the development of a field-based screening tool aimed at identifying those at high risk of developing OA after a youth sport-related knee injury to target with a secondary prevention intervention should include the assessment of: injury type (ACL rupture and/or meniscal injury), knee surgery history (ACLR and/or meniscus), knee-related quality of life and a functional test with strength demands (eg, triple-hop). Although only treated as a covariate in this investigation, obesity26 is an accepted modifiable risk factor for post-traumatic OA and therefore screening for higher BMI or per cent fat mass would also be appropriate.

Strengths and limitations

The strengths of this investigation include the incorporation of healthy controls matched on age, sex and sport exposure, a broad definition of knee injury and confirmation of intra-articular knee injury by a physiotherapist or physician at the time of injury, which minimises misclassification bias. Due to convenience sampling, it is plausible that previously injured youth experiencing more symptoms and functional restrictions were more likely to participate than those without complaints. However, as the range of scores for the self-reported and functional outcomes in participants previously injured includes scores associated with no pain, no symptoms or no functional restrictions, we are confident that this is not the case. As the surgical reports of previously injured participants who underwent ACLR were not available for review, it is impossible to know if concomitant meniscal lesions were repaired or excised, which would impact the development of MRI-defined OA. With that said, it is common practice to repair concomitant meniscal lesions at the time of ACLR in youth and young adults. Given that the MRI studies are from normal to nearly normal knees in youth and young adults, the reliability for some MOAKS subscales was, as expected, lower than what has been previously reported in patients with established structural damage from OA. As a single board-certified musculoskeletal fellowship-trained radiologist assessed all MRI studies, it is likely that any measurement bias is systematic in nature. A larger sample size would allow for the consideration of interactions between the variables within the multivariable models and the inclusion of other potential explanatory factors, which could shed further light on the relationship between MRI-defined OA and each outcome. A larger sample size would also result in a more precise estimate and narrower 95% CI for the OR reported. Finally, participants in this study represent an active youth sport population, symptomatology; therefore, findings may not be generalisable to less active youth or an older post-traumatic population with OA.


Three to 10 years history of youth sport-related intra-articular knee injury is associated with an increased OR of MRI-defined OA, and MRI features associated with future OA exist in young adults who have sustained a range of severity of knee injury and are not exclusive to those who have sustained ACL tears and meniscal lesions. Finally, MRI-defined OA has a linear association (independent of injury history) with self-reported knee-related quality of life but not knee pain, symptoms or function in daily living, knee muscle strength and triple-hop or YBT. These findings suggest that 3–10 years following knee injury, structure may not dictate clinical symptomatology (ie, self-reported pain, symptoms or function), muscle strength or function but may influence quality of life. A definition of preradiographic OA and/or screening tools to assess who to target with secondary OA prevention strategies should consider structural changes, clinical symptomatology, knee muscle strength or function and quality of life.

What are the findings?

  • Structural findings consistent with osteoarthritis (OA) are not unique to individuals who have torn their ACL and/or injured their menisci but are also present in individuals 3–10 years after a patellofemoral subluxation/dislocation or knee collateral ligament sprain.

  • MRI-defined OA 3–10 years following knee injury was not found to be independently associated with self-reported knee pain, symptoms or function in daily living, knee muscle strength or dynamic balance but was independently associated with knee-related quality of life, suggesting that structure may not dictate clinical symptomatology.

  • Further research is required to determine what combination of self-reported clinical, functional and structural outcomes is best for detecting individuals that would benefit from secondary OA prevention strategies after a youth sport-related knee injury.

How might it impact on clinical practice in the future?

  • MRI is not sufficient for a clinical diagnosis of OA.

  • Clinical screening tools aimed at identifying individuals at high risk of developing OA after a youth sport-related knee injury to target with secondary prevention interventions should consider clinical symptomatology, knee muscle strength or function and quality of life alongside structural changes.

  • Secondary OA prevention strategies should be expanded to include individuals that have not sustained an ACL and/or meniscal injury.


The authors would like to acknowledge the assistance of research coordinators Gabriella Nasuti, Jamie Rishaug and Lisa Loos and the numerous Sport Injury Prevention Research Centre research assistants/students as well as the participants.



  • Contributors CAE, LJW, JLJ and JLW conceived and designed the study and obtained funding. JLW and CAE coordinated the study and managed all aspects including collection and assembly of the data. JLJ oversaw all diagnostic imaging. AN-A, CAE and JLW planned the analyses. JLW conducted all analyses and wrote the first draft of the manuscript. All authors contributed to the interpretation of the data and critical revision and approval of the final manuscript.

  • Funding The Alberta Youth Prevention of Early Osteoarthritis cohort is funded by the Canadian Institute of Health Research (MOP 133597), the Alberta Osteoarthritis Team supported by Alberta Innovates Health Solutions (AIHS) and the Alberta Children’s Hospital Foundation. The Sport Injury Prevention Research Centre is supported by an International Olympic Committee Research Centre Award. JLW and CMT were awarded AIHS Clinician Fellowships to support this cohort study.

  • Disclaimer The funders had no role in any part of the study or the decision to publish. All authors had full access to the data and take responsibility for data integrity and the accuracy of the analyses. The senior author (CAE) affirms that this manuscript is an honest, accurate and transparent account of the study being reported, that no important aspects of the study were omitted and that any discrepancies from the planned study have been explained.

  • Competing interests CAE is funded through a Chair in Pediatric Rehabilitation (Alberta Children’s Hospital Foundation). LJW has a consultation arrangement with Eli Lilly. The remaining authors have nothing to disclose.

  • Patient consent Obtained.

  • Ethics approval Conjoint Health Research Ethics Board, University of Calgary, Canada.

  • Provenance and peer review Not commissioned; externally peer reviewed.