Objective Varus and valgus alignment are associated with progression of knee osteoarthritis, but their role in incident disease is less certain. Radiographic measures of incident knee osteoarthritis may be capturing early progression rather than disease development. The authors tested the hypothesis: in knees with normal cartilage morphology by MRI, varus is associated with incident medial cartilage damage and valgus with incident lateral damage.
Methods In MOST, a prospective study of persons at risk of or with knee osteoarthritis, baseline full-limb x-rays and baseline and 30-month MRI were acquired. In knees with normal baseline cartilage morphology in all tibiofemoral subregions, logistic regression was used with generalised estimating equations to examine the association between alignment and incident cartilage damage adjusting for age, gender, body mass index, laxity, meniscal tear and extrusion.
Results Of 1881 knees, 293 from 256 persons met the criteria. Varus versus non-varus was associated with incident medial damage (adjusted OR 3.59, 95% CI 1.59 to 8.10), as was varus versus neutral, with evidence of a dose effect (adjusted OR 1.38/1° varus, 95% CI 1.19 to 1.59). The findings held even excluding knees with medial meniscal damage. Valgus was not associated with incident lateral damage. Varus and valgus were associated with a reduced risk of incident lateral and medial damage, respectively.
Conclusion In knees with normal cartilage morphology, varus was associated with incident cartilage damage in the medial compartment, and varus and valgus with a reduced risk of incident damage in the less loaded compartment. These results support that varus increases the risk of the initial development of knee osteoarthritis.
Statistics from Altmetric.com
In persons with established knee osteoarthritis, varus alignment is associated with subsequent progression of medial tibiofemoral osteoarthritis and valgus with progression of lateral osteoarthritis.1,–,6 Varus alignment has also been found to be associated with incident radiographic knee osteoarthritis in the two cohort studies evaluating this4 ,6 but not in a case–control study;7 the association between valgus alignment and incident radiographic osteoarthritis was borderline4 or not evident.6 The role of alignment in the initial development of knee osteoarthritis is less certain, in large part due to inherent limitations of radiography to define incident osteoarthritis.
The established radiographic definition of incident knee osteoarthritis (Kellgren/Lawrence (K/L) grade ≥2, presence of definite osteophytes) cannot fully capture the magnitude of the effect of a risk factor like varus alignment, which potentially stresses one compartment (medial tibiofemoral) while off-loading the other (lateral) compartment. Osteophyte development is an early but not compartment-specific event in osteoarthritis; this definition cannot distinguish incident medial versus lateral osteoarthritis. Also, radiography is insensitive to early cartilage damage. Cartilage damage may already be present in knees defined as ‘at risk’ of incident radiographic osteoarthritis, ie, knees graded K/L 0 or 1,8 particularly in individuals at higher risk of developing knee osteoarthritis, begging the question: does a radiographic measure of incident knee osteoarthritis capture the initial development of disease or worsening of early, established disease? In contrast to radiography, MRI allows the identification of knees that are free of any cartilage damage at baseline for prospective study and provides a compartment-specific outcome measure.
The load-bearing axis, femoral head centre to ankle joint centre, passes medial to knee centre in a varus knee, increasing force across the medial compartment and lateral to knee centre in a valgus knee, increasing force across the lateral compartment. In a healthy knee during normal gait, load distribution is not equal between medial and lateral compartments: 70% of load passes through the medial compartment, primarily due to an external knee adduction moment.9 ,10 With greater varus, the proportion of load distributed medially increases further.11 ,12 With greater valgus, load distribution shifts from greater medial, to equal, to greater lateral (with more severe valgus).13,–,15 It is likely that in a healthy knee, the mechanical impact of varus on the medial compartment exceeds that of valgus on the lateral compartment.
We identified knees with normal cartilage morphology by MRI in all tibial and femoral subregions at baseline to test the hypothesis: varus alignment is associated with incident medial cartilage damage and valgus alignment is associated with incident lateral damage.
MOST is an observational cohort study of incident and progressive knee osteoarthritis in community-dwelling individuals (Iowa City and Birmingham), aged 50–79 years, recruited via mailings and community outreach. Participants were required to have symptomatic knee osteoarthritis or characteristics that placed them at increased risk of developing it.16 Exclusion criteria were bilateral knee replacement or plan for this within a year, inability to walk without another person or walker, serious health condition, inflammatory arthritis, dialysis, malignancy, plan to move away. The study protocol was approved by each site's institutional review board. All participants provided written, informed consent.
Measurement of varus–valgus alignment, laxity, weight and height
Alignment was assessed from baseline full-limb radiographs, acquired as previously described.1 One anteroposterior radiograph of both limbs was obtained, using a 51×14 in graduated grid cassette (Iowa City) and a CR-based system of overlapping cassettes and simultaneously exposed subimages forming a stitched image (Birmingham).
Alignment (hip–knee–ankle angle) was measured as the angle at the intersection of lines connecting femoral head and intercondylar notch centres and connecting ankle talar surface centre and tibial interspinous sulcus base. Image analysis17 was completed by one of three trained readers using a customised program (Surveyor 3; OAISYS Inc, Kingston, Ontario, Canada), blinded to all other data. In a reliability study of 200 full-limb pairs assessed by these three readers, the inter and intrareader intraclass correlation coefficients were 0.95 and 0.96.18 In analyses, varus was defined as 178° or less, valgus as 182° or greater and neutral as 179–181°.
Weight (kg) without shoes or heavy clothes or jewellery was measured on a standard balance beam scale and height without shoes at peak of inhalation using a Harpenden stadiometer (Holtain, Pembrokeshire, UK). Body mass index (BMI) was calculated as weight in kilograms divided by height in square metres. Medial–lateral laxity (°) was measured using a protocol and device previously described,19 consisting of a bench and arc-shaped track, and providing thigh and ankle immobilisation, a stable knee flexion angle and fixed medial and lateral load.
Knee x-ray acquisition and assessment
Knee radiographs were acquired using the posteroanterior fixed-flexion weight-bearing protocol,20 with knees flexed to 20–30° and feet internally rotated 10° using a plexiglass positioning frame (SynaFlexer SynarcTM, San Francisco, California). Knees were imaged together on 14×17 film with a 72 in film-to-focus distance. An experienced rheumatologist and musculoskeletal radiologist independently assessed each film for K/L grade. Disagreement was adjudicated by a panel of three readers. Weighted κ for agreement between the readers was 0.79 for K/L grade.
MRI acquisition and measurements
At baseline and 30-month follow-up, bilateral knee MRI were obtained with a 1.0-T dedicated system (ONI MSK Extreme; GE Healthcare, Waukesha, Wisconsin, USA) with a circumferential extremity coil by using fat-suppressed fast spin-echo intermediate-weighted sequences in the sagittal (repetition time ms/echo time ms, 4800/35; section thickness, 3 mm; intersection gap, 0 mm; sections, 32; matrix, 288×192; signals acquired, two; field of view, 140 mm2; echo train length, eight) and axial (4680/13; section thickness, 3 mm; intersection gap, 0 mm; sections, 20; matrix, 288×192; signals acquired, two; field of view, 140 mm2; echo train length, eight) planes and a short τ inversion recovery sequence in the coronal plane (6650/15; inversion time, 100 ms; section thickness, 3 mm; intersection gap, 0 mm; sections, 28; matrix, 256×192; signals acquired, two; field of view, 140 mm2; echo train length, eight).
Two musculoskeletal radiologists (FR and AG), blinded to all other data, evaluated the images using the whole-organ MRI score (WORMS).21 Paired baseline and follow-up images were read, with the chronological order known.22 Cartilage signal intensity and morphology were scored according to WORMS from 0 to 6 (depending on depth and extent of cartilage loss) in five subregions each in the medial and lateral tibiofemoral compartments, for a total of 10 tibiofemoral subregions. Meniscal status was graded from 0 to 4 in the anterior horn, body and posterior horn of each meniscus, defining tear as a WORMS score greater than 0 in one or more segments. In addition to the WORMS, extrusion of each meniscal body was scored on the coronal image as 0–2, defining extrusion as a score greater than 0.23
Definition of outcome
While MRI to score cartilage damage has been extensively validated, the best definition(s) of incident OA has not been established. The transition essential to our hypothesis, which deals with the initial development of cartilage damage, was from normal to the lowest score with an unequivocal cartilage lesion. All knees in the analysis sample had a cartilage score of 0 (normal cartilage signal intensity and morphology) in all tibiofemoral subregions at baseline. Incident medial cartilage damage was defined as the development of a cartilage morphology score of 2 or greater (2=solitary partial thickness defect) in one or more medial subregions by 30 months. Incident lateral cartilage damage was defined as the development of a cartilage morphology score of 2 or greater in one or more lateral subregions by 30 months.
The analysis sample included only knees with normal baseline tibiofemoral cartilage signal intensity and morphology in both the medial and lateral tibiofemoral compartments (as defined above). All statistical analyses were knee based. Logistic regression models with generalised estimating equations were used to account for potentially correlated observations between knees from the same person. Using these models, we examined the association between baseline alignment in each knee and the subsequent development of incident cartilage damage at 30 months (as defined above).
Inferences from the analyses were based on the following three model-based comparisons for the outcome of incident medial cartilage damage by 30 months: varus knees versus all other knees (or non-varus) as reference; varus knees versus neutral knees as reference; and varus alignment considered as a continuous variable in the models. A similar model-based approach was used to evaluate the association of valgus alignment and incident lateral cartilage damage.
All analyses were adjusted for age (continuous), gender, BMI (continuous) and laxity (continuous), and then further adjusted for meniscal tear and extrusion (in the compartment of interest). Results from each model are reported as adjusted OR with an associated 95% CI that excludes one representing a statistically significant association. In secondary analyses, the severity of varus and valgus alignment were each analysed as continuous variables. Because the knees with MRI data were a subset of all knees from persons enrolled in MOST, we also conducted sensitivity analyses to address the potential for selection bias due to the MOST substudy sampling mechanisms (that included several case–control and cohort samples to address specific research questions). In these sensitivity analyses, we excluded MOST substudies that selected knees for MRI reading based on criteria that may have been linked to either alignment or to our outcome, ie, we excluded the knees from case–control studies of incident radiographic knee osteoarthritis and incident knee symptoms. Analyses were performed using SAS software version 9.2.
As shown in figure 1, by the 30-month follow-up, of the 3026 persons enrolled, 33 had died, 24 could not be reached and 2969 were reached by phone, of whom 2713 completed the visit. Reasons for not completing the 30-month visit were: too busy, 77 persons; health problems, 70; caregiving responsibilities, 31; deceased, 30; clinic distance, 21; had moved, 20; study dissatisfaction, 19; unable to contact, 16; refused to give reason, eight; personal problems, seven; and other reasons, 14. Those not completing the 30-month visit did not differ in age, gender, or dominant knee alignment distribution (32% neutral, 47% varus, 21% valgus) but had a higher BMI (32.0±6.9 (SD) vs 30.6±5.8) than those who completed this visit. The steps to derive the analysis sample, 293 knees (from 256 persons) without any cartilage damage in any tibiofemoral subregion at baseline, are delineated in figure 1.
The 256 persons had a mean age of 60.0 years (±7.5), a mean BMI of 28.6 kg/m2 (±4.5) and included 172 (67.2%) women. Eighty-two (32.0%) of the 256 had a BMI of 30 kg/m2 or greater. Of the 293 knees (258 K/L 0, 25 K/L 1 and 10 K/L 2), 128 (43.7%) were neutral, 110 (37.5%) were varus and 55 (18.8%) were valgus. Knee characteristics stratified by alignment are shown in table 1.
The distribution of 30-month outcomes by baseline alignment category is shown in figure 2. As shown in table 2, varus versus non-varus was associated with incident medial cartilage damage (adjusted OR 3.59, 95% CI 1.59 to 8.10). Varus versus neutral alignment was also associated with incident medial damage. Greater varus was associated with incident medial damage (adjusted OR 1.38/1° varus, 95% CI 1.19 to 1.59). Sensitivity analyses yielded similar results, ie, adjusted effect estimates that ranged from within 1% to within 8% of the effect estimates from the fully adjusted analyses provided in the furthest right column of table 2. As shown in table 3, neither valgus versus non-valgus nor valgus versus neutral was associated with incident lateral damage. Greater valgus was associated with the outcome, but this was not significant after adjustment for meniscal tear and extrusion. Varus was associated with a reduced risk of incident damage in the off-loaded, lateral compartment (adjusted OR 0.75/1°, 95% CI 0.59 to 0.95) and valgus with a reduced risk of incident medial damage (adjusted OR 0.72/1°, 95% CI 0.63 to 0.83). Results of the fully adjusted models of tables 2 and 3 were minimally altered by adjusting for the baseline to 30-month change in BMI.
To explore whether an alignment effect requires existing meniscal damage, we repeated these analyses in knees without meniscal tear or extrusion. Excluding knees with baseline medial meniscal tear or extrusion left 205 knees (78 varus, 40 valgus, 87 neutral), of which 16 (7.8%) had incident medial cartilage damage by 30 months. In these 205 knees, greater varus was associated with incident medial cartilage damage, adjusting for age, gender, BMI and lateral laxity (adjusted OR 1.37/1° varus, 95% CI 1.11 to 1.68). Excluding knees with baseline lateral meniscal tear or extrusion left 274 knees (107 varus, 49 valgus, 118 neutral), 10 (3.6%) with incident lateral damage. In these 274 knees, greater valgus was not associated with incident lateral damage (adjusted OR 1.14/1° valgus, 95% CI 0.89 to 1.46).
In knees without MRI evidence of cartilage damage in any tibiofemoral subregion, varus alignment at baseline was associated with an increased risk of incident medial tibiofemoral cartilage damage over the subsequent 30 months, whether compared with knees without varus alignment as a whole or specifically with neutral knees. Greater varus angle was associated with a greater risk of incident medial damage and a reduced risk of incident lateral damage. Valgus was not associated with an increased risk of lateral tibiofemoral cartilage damage, compared either with knees without valgus as a whole or with neutral knees. Greater valgus angle was not associated with a greater risk of incident lateral damage but was associated with a reduced risk of incident medial damage.
Previous studies of osteoarthritic knees support a relationship between varus and subsequent medial progression and between valgus and lateral progression, using radiographic and MRI outcome measures.1,–,6 However, a key question has not as yet been addressed: does varus or valgus alignment preceding osteoarthritis influence the risk of osteoarthritis development? Radiographic studies employing the established definition of incident osteoarthritis (ie, new development of K/L ≥2 in knees K/L 0 or 1 at baseline) cannot answer this question, as K/L 0–1 knees may already have osteoarthritis, given the insensitivity of x-ray to early disease.8 The possibility that a K/L 0 or 1 knee has cartilage damage may be greater in persons at higher risk of Osteoarthritis (such as in MOST or the osteoarthritis initiative) than in the general population. MRI affords an excellent means of identifying knees with normal cartilage morphology at baseline. Only knees with normal cartilage morphology in every tibiofemoral subregion at baseline, both medial and lateral, were eligible for this study. This approach provided an outcome measure to capture the initial development of cartilage damage, the transition at the core of the difficult question about the impact of alignment preceding osteoarthritis. With only two time points, it was not possible for us to evaluate whether any alignment effect on subsequent cartilage damage is via changes to other tissues such as subchondral bone or synovium, an interesting question for future longer follow-up studies.
These results, an association of varus with an increased risk of incident medial cartilage damage and a reduced risk of lateral damage, support the theory that the mechanism of action of varus relates to load distribution. The low frequency of valgus and of incident lateral damage in the current study reduced our power to detect their relationship. On the other hand, a stronger finding for varus than for valgus is not surprising. Due to a stance phase knee adduction moment, greater load passes medially than laterally even in neutral, healthy knees.9 ,10 The adduction moment magnitude increases as varus increases.11 The adduction moment magnitude predicted knee osteoarthritis progression24; it may lie in the causal pathway between varus and knee osteoarthritis progression. Varus further increases total load passing medially.11 ,12 Although valgus alignment is associated with an increase in lateral compartment peak pressures,13 more load is still borne medially until more severe valgus is present.14 ,15 In keeping with a less potent effect of valgus, cohort studies have found that varus but not valgus (vs neutral) increased the risk of incident radiographic osteoarthritis.4 ,6
Our study has limitations. Individuals in the MOST cohort without knee osteoarthritis were at higher risk of developing it. However, the population at higher risk of developing knee osteoarthritis is of public health importance and the deliberate focus of the MOST study design.16 The size of this group, already large, will grow as the ageing segment expands, and it is essential to understand the impact of alignment in them. The MOST substudy sampling design to select knees for MRI reading may limit generalisability to the entire MOST cohort; however, our sensitivity analyses did not reveal evidence of bias in our findings. As noted above, the numbers of valgus knees and knees with incident lateral cartilage damage were low. Of note, image data from nearly 2000 knees were required to derive the eligible knees. To address the questions posed, it was essential to adhere to strict criteria regarding baseline cartilage morphology.
These results are not solely of academic importance. The source of malalignment predating knee osteoarthritis may be genetic, developmental or traumatic. In an individual with a varus knee, it is possible that an intervention that improves medial to lateral load distribution could help to delay the onset of osteoarthritis; at this early point, such a strategy may be more powerful than when osteoarthritis has established its presence in a knee. These results point to the importance of continuing to develop, refine and test non-invasive, simple and inexpensive interventions (eg, orthoses, gait modifications) for varus knees.
In conclusion, in knees without MRI evidence of cartilage damage in any tibiofemoral subregion, varus alignment at baseline was associated with an increased risk of new cartilage damage in medial subregions over the subsequent 30 months, whether compared with knees without varus or specifically with neutral knees. A dose effect was present, whereby greater varus angulation was associated with a greater risk of incident medial damage and a reduced risk of incident lateral damage. These results suggest that varus alignment is a risk factor for incident cartilage damage and provide further evidence that varus alignment increases the risk of the initial development of knee osteoarthritis.
Funding Support for this project comes from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, R01 HD43500 (LS), and the National Institute on Ageing, U01 AG18820 (DF), U01 AG18832 (JT), U01 AG18947 (CEL) and U01 AG19069 (MN).
Competing interests None.
Ethics approval The study protocol was approved by each site's institutional review board.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
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.