Knee alignment does not predict incident osteoarthritis: The Framingham osteoarthritis study

Authors


Abstract

Objective

To examine the relationship of knee malalignment to the occurrence of knee osteoarthritis (OA) among subjects without radiographic OA at baseline to determine whether malalignment is a risk factor for incident disease or simply a marker of increasing disease severity.

Methods

We selected 110 incident tibiofemoral (TF) OA case knees (76 subjects) and 356 random control knees (178 subjects) from among participants in the Framingham Osteoarthritis Study. Case knees did not have OA at baseline (1992–1994 examination) but had developed OA (Kellgren/Lawrence grade ≥2) at followup (2002–2005 examination) (mean of 8.75 years between examinations). Control knees did not have OA at baseline. Standardized digital radiographs of the fully extended knee with weight-bearing were read using a standard protocol and eFilm viewing software. We measured the anatomic axis, the condylar angle, the tibial plateau angle, and the condylar tibial plateau angle. The interobserver intraclass correlation coefficient (ICC) ranged from 0.93 to 0.96 and the intraobserver ICC from 0.94 to 0.97. In a knee-specific analysis, we examined the relationship of each alignment measurement to the risk of TF OA using generalized estimating equations, adjusting for age, sex, and body mass index (BMI). We used the same approach to assess the association between each alignment measurement and the risk of medial TF OA.

Results

Subjects in the case population were older and had a higher BMI than the controls. The alignment values were normally distributed and were not different between the cases and the controls. After adjustment for age, sex and BMI, there was no significant increase in incident OA in the highest quartile compared with the lowest quartile category for any of the alignment measures (P for trend for anatomic axis and condylar tibial plateau angle was 0.83 and 0.80, respectively). Similar results were also observed for medial compartment OA.

Conclusion

We found that baseline knee alignment is not associated with either incident radiographic TF OA or medial TF OA. These results suggest that malalignment is not a risk factor for OA, but rather is a marker of disease severity and/or its progression.

Osteoarthritis (OA) is widely believed to result from local mechanical factors acting within the context of systemic susceptibility. Previous studies have demonstrated that malalignment is a potent predictor of disease progression in knee OA (1). However, these observations have only been made in populations with preexisting radiographic disease. While frontal plane malalignment is clearly associated with increasing structural damage in the knee joint, it is still unclear whether the observed malalignment is a consequence or a cause of worsening disease (2). In the effort to better understand the etiology of knee OA and develop effective preventative interventions, it is essential that we clarify the role of knee malalignment in both the initial occurrence of the disease and its subsequent progression.

There is good reason to believe that OA susceptibility may be affected by differences in alignment. In theory, any shift from a neutral or collinear alignment of the hip, knee, and ankle affects load distribution at the knee (3). The load-bearing mechanical axis is traditionally represented on radiographs by a line drawn from the center of the femoral head to the center of the tibial spines and a second line from the center of the ankle talus to the center of the tibial spines. In a varus malaligned knee, this load-bearing line passes medial to the center of the knee, and an adduction moment arm is created, which increases compressive force across the medial compartment. In a valgus malaligned knee, the load-bearing axis passes lateral to the knee center, and the resulting abduction moment arm increases force across the lateral compartment (3). More recently, the intraarticular alignment of the femoral condyles and the tibial plateau has been proposed as a more valid indicator of how contact forces are distributed across the joint surfaces of the knee during weight bearing activities (4, 5). Under conditions of normal alignment, the medial compartment absorbs 60–70% of the compressive force across the knee during weight bearing (6). This fact has been offered as a possible explanation for why OA affects the medial tibiofemoral compartment much more often than the lateral compartment in both men and women (7).

In the presence of existing knee OA, abnormal anatomic alignment is associated with accelerated structural deterioration in the compartment under greatest compressive stress. Varus malalignment has been shown to lead to a 4-fold increase in focal medial knee OA progression, while valgus malalignment has been shown to predispose to a 2–5-fold increase in lateral OA progression (1, 8). Malalignment also appears to be a major determinant of the size, location, and progression of bone marrow lesions, with subsequent impact on the rate of cartilage loss (9). However, these findings are from studies of persons with preexisting disease. While frontal-plane knee malalignment may be a potent risk factor for disease progression and/or a useful clinical marker of increasing disease severity, it has not yet been shown to be a risk factor for incident knee OA. Therefore, we performed a nested case–control study among participants in the Framingham Osteoarthritis Study in order to evaluate whether baseline malalignment (both anatomic axis and condylar plateau angle) was associated with subsequent incident radiographic knee OA ∼9 years later.

SUBJECTS AND METHODS

The Framingham OA Study is a population-based cohort study of risk factors for OA. The Framingham OA Study includes a cohort of offspring of the original Framingham Heart Study, which was begun in 1948 (10). As part of a study on the inheritance of OA, participants in the Framingham Offspring Cohort were originally examined between 1992 and 1994 (examination 5 of the Framingham Heart Study). All surviving members of this group were contacted by telephone and invited to participate in a more recent examination from 2002 to 2005 (mean interval between visits 8.75 years). A validated survey instrument (11), supplemented by questions about medication use, was used to exclude patients with rheumatoid arthritis.

Radiographic assessments.

During the Framingham Offspring Cohort examination 5 visit (conducted during 1992–1994), a radiograph of both knees in full extension with weight-bearing was obtained, using a standardized protocol that included outlines of the feet in order to keep the rotation of the knee the same at followup evaluations. Knee radiographs were obtained at 0° and at 6° caudad, and the better of the 2 views (based on the optimal superimposition of the anterior and posterior margins of the medial tibial plateau) was selected for comparison with findings at the followup examination.

After the subjects completed their followup examination, both baseline and followup radiographs were read independently by 2 study readers, one a bone and joint radiologist (PA), and the other a rheumatologist (BS). Knee radiographs were read in a paired manner, unblinded with regard to chronological sequence. Readers graded each knee according to the Kellgren/Lawrence (K/L) method (12) at both time points and evaluated the individual features and whether these features had changed over the followup period. Each knee was scored on a 0–3 scale for the presence of osteophytes and joint space narrowing (JSN) using the Osteoarthritis Research Society International atlas (13).

For the K/L grade and its change, we used readings from the bone and joint radiologist (PA). The adjudication sessions were chaired by one of us (DTF). If there was a difference in K/L grade based on osteophytes that led to a difference in the assignment of either prevalent or incident OA status, this was adjudicated. All adjudicated readings were arrived at by consensus of the readers. Disagreements on changes in the joint space were also adjudicated if the two readers disagreed and if adjudication of joint space change altered the appropriateness of the K/L score. For the bone and joint radiologist, the intrareader kappa value for the K/L grade was 0.82; the interreader kappa value was 0.74 (P < 0.001 for both comparisons).

Selection of case and control groups.

Subjects eligible for incident radiographic tibiofemoral (TF) OA were participants in the Framingham Offspring Cohort Study who had no radiographic TF OA (K/L grade <2) in both knees at baseline visit (during 1992–1994) (12). During the study period, 110 knees (from 76 subjects) developed incident radiographic TF OA (K/L grade ≥2) at the followup visit (during 2002–2005). These knees served as the case group. The control group was randomly selected from among subjects who were eligible for incident radiographic TF OA and were in the same age and body mass index (BMI) range as the case group. A total of 356 knees (from 178 subjects) served as controls.

In a subgroup analysis aimed at determining the risk of incident medial compartment OA, we further identified 80 case knees with isolated medial TF OA. Medial TF OA was defined as medial joint space narrowing greater than lateral joint space narrowing and a K/L grade ≥2 (including a definite osteophyte).

Measurements of alignment.

We measured knee alignment among all case and control knees. The radiographs were digitized so that imaging software (eFilm Workstation version 2.0.0; Merge, Milwaukee, WI) could be used to compose reference lines and calculate the following angular measures (see Figure 1): the anatomic axis alignment (in degrees from 180°), the femoral condylar angle (in degrees from 90°), the tibial plateau angle (in degrees from 90°), and the condylar tibial plateau angle (in degrees between the femoral condylar and tibial plateau lines).

Figure 1.

Measures of knee alignment in a varus knee. The anatomic axis (AA) is the angle between lines drawn from the midpoint of the femur through the tibial spines and from the tibial spines to the midpoint of the tibia. The condylar angle (CA) is the angle between the mechanical or anatomic axis line of the femur and a line tangent to the femoral condyles. The tibial plateau angle (PA) is the angle between the mechanical or anatomic axis line of the tibia and a line tangent to the tibial plateau. The condylar tibial plateau (CP) angle is the angle between the above-mentioned tangent lines.

Our assessment of the anatomic axis is consistent with that described by Kraus et al (14). The anatomic axis was defined as the angle formed by the intersection of 2 lines originating from points bisecting the femur and tibia and converging at the center of the tibial spine tips. The origin of these lines was 10 cm from the knee joint surfaces when included in the field of view on the radiograph.

Given the possibility of discrepant findings between condylar tibial plateau alignment and the more traditional anatomic axis measure of knee alignment, we were careful to assess both. Varus angles were assigned negative values, and valgus angles were assigned positive values. All measurements were taken by a single reader (WFH) who was blinded to the study question and the outcome status of the subject. To evaluate for reader drift, we reassessed intrarater reliability by inserting 1 original reliability radiograph for every 10 new radiographs read. The intrarater intraclass correlation coefficient (ICC) ranged from 0.94 to 0.97, and the interrater ICC ranged from 0.93 to 0.96 for the different angular measures.

Statistical analysis.

We divided anatomic axis alignment, femoral condylar angle, and condylar tibial plateau angle into 4 categories according to the quartile distribution of each measurement among all subjects. Knees in the lowest category of specific alignment measurement were used as the referent group. We examined the relationship of each alignment measurement to the risk of incident TF OA, adjusting for age, BMI and sex.

Previous studies have suggested that incident OA develops more frequently in knees that already have doubtful OA (K/L grade 1) (15, 16). Similarly, increasing malalignment has been associated with increasing disease severity (17). Rather than just assuming that the baseline K/L grade did not effect either the subsequent risk of progression or the alignment, we stratified the knees according to their baseline K/L score (0 versus 1) and obtained a pooled effect estimate of each alignment measurement on the risk of incident knee OA using a conditional logistic regression model with generalized estimating equations to control for correlation between two knees in these models. The variability of the tibial plateau angle measurement was small, and most values were close to the mean; therefore, we divided this measurement into tertile groups according to the distribution among all participants. We used the same approach to assess the relationship between each alignment measurement and incident TF OA of the medial compartment. We assessed for both linear and U-shaped trends given the potential for risk in both valgus and varus directions in the incident TF OA analysis and tested for linear trends for the analysis of incident TF OA in the medial compartment.

RESULTS

Of 1,705 subjects seen at the initial examination, 1,279 (75.0%) obtained knee radiographs as part of the followup examination, with an average (±SD) interval of 8.75 ± 1.0 years. Among the 2,259 knees undergoing both baseline and followup examination, we identified 110 cases of incident TF OA (76 subjects). We randomly selected 356 control knees (178 subjects) from the remaining participants in the Framingham Osteoarthritis Study.

The characteristics of the study population are displayed in Table 1. On average, subjects in the case population were older, had a higher BMI, and were more likely to have had a prior knee injury than subjects in the control population. All angular measures were similar between the two groups, with the exception of the tibial plateau angle. The case knees had a smaller tibial plateau angle than the control knees (P = 0.009); however, the alignment measurement derived from this, the condylar tibial plateau angle, was not significantly different between the two groups.

Table 1. Characteristics of the study population
 Cases (n = 110 knees)Controls (n = 356 knees)
  • *

    K/L = Kellgren/Lawrence.

  • Positive numbers indicate valgus angles; negative numbers indicate varus angles.

Age, mean ± SD (range) years55.2 ± 8.2 (32–74)52.7 ± 8.1 (32–73)
Body mass index, mean ± SD30.5 ± 5.627.1 ± 4.7
% female53.653.6
% of knees with K/L score of 1 at baseline*35.57.3
% with history of knee injury23.99.9
% who contributed 2 knees to the analysis44.798.9
Measures of alignment, mean ± SD (range) degrees  
 Anatomic axis2.0 ± 3.0 (−7 to 10)2.2 ± 3.2 (−7 to 10)
 Condylar angle5.4 ± 3.0 (−6 to 11)5.2 ± 3.0 (−3 to 13)
 Tibial plateau angle−1.1 ± 1.2 (−5 to 3)−0.8 ± 1.0 (−4 to 4)
 Condylar tibial plateau angle−2.4 ± 1.8 (−6 to 3)−2.3 ± 1.8 (−7 to 5)

At followup, most of the incident knee OA cases were K/L grade 2 (73%). Of the remaining case knees 25% were K/L grade 3, and <3% were K/L grade 4. Of the control sample (without OA at baseline), the majority of knees at followup were K/L grade ≤2 (91.8%), while a much smaller proportion were K/L grade 2 (5.9%), grade 3 (1.7%), or grade 4 (0.6%).

The relationship of each knee alignment measurement to the risk of incident knee OA is presented in Table 2. The risk of incident knee OA did not increase with either increasing varus or increasing valgus anatomic axis malalignment. Using the lowest quartile as a reference group, the odds ratio was ∼1.1 in each of the second, third, and fourth quartiles of anatomic alignment (P = 0.83 for trend) after adjusting for age, sex, and BMI. Similarly, the relative odds of incident knee OA for the second, third, and fourth quartiles of the condylar tibial plateau angle measured 0.94, 0.71, and 1.12 (P = 0.80 for trend) after adjusting for age, sex, and BMI.

Table 2. Relationship between alignment measurements and prevalence of incident osteoarthritis (adjusted for age, sex, and body mass index), by quartile of alignment*
 Quartile of alignmentP for trend
1 (reference group)234
  • *

    Positive numbers indicate valgus angles; negative numbers indicate varus angles. Tibial plateau angle values were divided into tertiles because of the small variability of the values, with most lying close to the mean. OR = odds ratio; 95% CI = 95% confidence interval.

Anatomic axis, degrees5–10 valgus3–4 valgus0–2 valgus1–7 varus 
 No. of case knees20323523 
 Prevalence21.123.927.620.9 
 Adjusted OR (95% CI)11.10 (0.56–2.19)1.07 (0.53–2.16)1.10 (0.52–2.30)Linear 0.83; U-shaped 0.84
Condylar angle, degrees−6 to 34 to 56 to 78 to 13 
 No. of case knees26223626 
 Prevalence20.021.029.024.3 
 Adjusted OR (95% CI)10.90 (0.44–1.84)1.31 (0.67–2.54)0.97 (0.47–1.99)Linear 0.97; U-shaped 0.29
Tibial plateau angle, degrees−5 to −2−1 to −10 to 4  
 No. of case knees393041  
 Prevalence36.418.520.8  
 Adjusted OR (95% CI)10.57 (0.31–1.05)0.68 (0.38–1.24) Linear 0.045; U-shaped 0.12
Condylar tibial plateau angle, degrees−7 to −4−3 to −3−2 to −2−1 to 5 
 No. of case knees29261639 
 Prevalence24.826.817.024.7 
 Adjusted OR (95% CI)10.94 (0.48–1.86)0.71 (0.33–1.52)1.12 (0.60–2.08)Linear 0.80; U-shaped 0.86

Of the 110 case knees with incident TF OA, 80 (73%) had medial TF OA at followup. Similar to the whole sample, the case subjects with medial TF OA were older and had a higher BMI than the control population (Table 3). The association between each angular measurement and the risk of incident medial OA is shown in Table 4. The odds ratios for the risk of incident medial knee OA for each quartile of the anatomic axis were 1.0, 1.33, 1.35, and 2.04 (P = 0.15 for trend) after adjusting for age, sex, and BMI. The odds of incident medial knee OA for each quartile of the condylar tibial plateau angle were 1.0, 0.78, 0.51, and 0.95, respectively (P = 0.78 for trend) after adjusting for age, sex, and BMI.

Table 3. Baseline characteristics of the study population with medial tibiofemoral osteoarthritis at followup
 Cases (n = 80 knees)Controls (n = 356 knees)
  • *

    K/L = Kellgren/Lawrence.

  • Positive numbers indicate valgus angles; negative numbers indicate varus angles.

Age, mean ± SD (range) years56.2 ± 7.7 (41–74)52.7 ± 8.1 (32–73)
Body mass index, mean ± SD30.6 ± 5.227.1 ± 4.7
% female46.353.6
% of knees with K/L score of 1 at baseline*37.57.3
Measures of alignment, mean ± SD (range) degrees  
 Anatomic axis1.6 ± 3.2 (−7 to 10)2.2 ± 3.2 (−7 to 10)
 Condylar angle5.4 ± 3.3 (−6 to 11)5.2 ± 3.0 (−3 to 13)
 Tibial plateau angle−1.2 ± 1.3 (−5 to 3)−0.8 ± 1.0 (−4 to 4)
 Condylar tibial plateau angle−2.6 ± 1.7 (−6 to 1)−2.3 ± 1.8 (−7 to 5)
Table 4. Relationship of alignment measurements to prevalence of incident medial tibiofemoral osteoarthritis (adjusted for age, sex, and body mass index), by quartile of alignment*
 Quartile of alignmentP for linear trend
1 (reference group)234
  • *

    Positive numbers indicate valgus angles; negative numbers indicate varus angles. Tibial plateau angle values were divided into tertiles because of the small variability of the values, with most lying close to the mean. OR = odds ratio; 95% CI = 95% confidence interval.

Anatomic axis, degrees5–10 valgus3–4 valgus0–2 valgus1–7 varus 
 No. of case knees11242421 
 Prevalence12.819.020.719.4 
 Adjusted OR (95% CI)11.33 (0.62–2.87)1.35 (0.59–3.10)2.04 (0.85–4.92)0.15
Condylar angle, degrees−6 to 34 to 56 to 78 to 13 
 No. of case knees21112622 
 Prevalence16.811.722.921.4 
 Adjusted OR (95% CI)10.45 (0.19–1.11)1.00 (0.47–2.11)0.85 (0.39–1.87)0.99
Tibial plateau angle, degrees−5 to −2−1 to −10 to 4  
 No. of case knees312425  
 Prevalence31.315.413.8  
 Adjusted OR (95% CI)10.61 (0.31–1.19)0.61 (0.31–1.21) 0.14
Condylar tibial plateau angle, degrees−7 to −4−3 to −3−2 to −2−1 to 5 
 No. of case knees25181027 
 Prevalence22.120.211.418.5 
 Adjusted OR (95% CI)10.78 (0.36–1.65)0.51 (0.21–1.23)0.95 (0.48–1.87)0.78

DISCUSSION

We found that baseline knee anatomic axis alignment was not associated with either incident TF OA or incident medial TF OA. Similarly, the condylar tibial plateau angle did not predict incident knee OA. The results suggest that malalignment may not be a primary risk factor for incident knee OA, but rather, a marker of disease severity and/or its progression.

Varus and valgus malalignment have been shown to increase the risk of medial and lateral knee OA progression (1). The effect of malalignment extends beyond its direct effect on cartilage since it has been shown in association with the structural breakdown of other knee tissues as well, such as the neighboring bone marrow (18). Certain site-specific factors in the local joint environment, such as tibiofemoral congruence, anterior cruciate ligament integrity, and meniscal degeneration and position also govern how load is distributed across the articular cartilage of a given joint. Previously, Cooke et al (19) suggested that the loss of joint space may account for much of the malalignment that has been observed in cases of knee OA. These observations were, however, never quantified. We recently conducted a cross-sectional analysis which suggested that multiple factors, including cartilage loss, meniscal degeneration and position, bone attrition, osteophytes, and ligament damage, can contribute to knee malalignment (2). This analysis was performed in a population of persons with preexisting OA, in whom there is greater variability in measures of alignment than that seen in the analysis presented herein. Understanding the role alignment plays in the etiology of OA is important, since the effect of standard risk factors for progression of knee OA, including obesity (18), quadriceps strength (20), laxity (20), and stage of disease (1, 8), are influenced by the presence or absence of malalignment.

In contrast to all previous studies, we examined the effect of malalignment on persons without knee OA at baseline. We would posit, based on the findings of this study, that abnormalities in frontal plane knee alignment are typically a consequence and not a primary cause of OA. While increasing structural damage within the knee may help perpetuate a vicious circle of events by bringing about malalignment, which in turn, can further strain already damaged tissue, the initial presence of malalignment does not appear to significantly increase the risk posed to healthy knee tissue. This is consistent with previous studies suggesting that knees with increasing disease severity are more malaligned (17) and that the association of alignment with progression differs according to the severity of disease (8). The absence of a relationship between malalignment and incident knee OA persists whether alignment is determined between bony segments (anatomic axis alignment) or adjacent joint surfaces (condylar tibial plateau angle).

This study has a number of limitations which may impair the interpretation of the study findings. OA develops slowly, and the development of radiographic signs can take many years. As a result, the number of incident cases of radiographic OA in our cohort may have been limited by the relatively short period of followup (mean of 8.75 years). In addition, radiographs are unable to detect the earliest indications of OA; thus, many participants in this sample, both cases and controls, may have had more subtle signs of disease that remained undetected by the radiographic examination protocol used in this study. The use of more sensitive imaging technology, such as magnetic resonance imaging, may reveal different results.

The mechanical axis measured on a full-limb radiograph is the gold standard method for assessing knee alignment. All measures of alignment in this study were taken from a “short” radiograph of only the knee, as opposed to a full-limb radiograph (14). Measures on the short radiograph do not capture proximal and distal anatomy, but do avoid unwanted pelvic radiation, high cost, and specialized equipment. The acquisition and measurement of the mechanical axis of the knee (defined as the angle formed by the intersection of a line from the center of the head of the femur to the center of the tibial spines and a second line from the center of the ankle talus to the center of the tibial spines) is technically difficult and requires a full-limb radiograph. However, a recent study demonstrated strong correlation between data obtained from full-limb measurements of the mechanical axis and short radiograph measurements of the anatomic axis (14). In that study, the anatomic axis was offset a mean 4.21° of valgus from the mechanical axis (3.5° in women and 6.4° in men). The high correlation suggests that the cumbersome full-length radiographs could be exchanged for the more commonly obtained, standard knee radiographs during the radiographic assessment of knee alignment. However, while they may be strongly correlated, there are differences, and adequate exploration of the relationship between alignment (mechanical axis) and incident OA should be pursued in a study where these measures are available.

Malalignment provides only a static impression of the expected uniplanar forces imparted on the joint during activity. The dynamics of gait involve multiple forces that stress the knee in several planes simultaneously. During the stance phase of gait, the ground reaction force acting at the foot passes medial to the axis of the knee joint. The perpendicular distance from the line of action of this force to the axis of the knee joint constitutes a moment arm for the application of an adduction moment to the knee (6). In a varus malaligned knee, the peak adduction moment is expected to increase, augmenting the load across the medial compartment (3). Yet, static measures of alignment cannot tell us with certainty whether any dynamic moments around the knee, including the adduction moment, are causative of incident knee OA. The adduction moment might also be affected by habitual postures during locomotion or by more distal malalignments, such as tibial or calcaneal varum. The possible role of distal malalignments in the etiology of knee OA is poorly understood and warrants a great deal more exploration.

Neither baseline anatomic axis alignment nor condylar tibial plateau angle is associated with incident TF OA. The absence of an association persists even when only medial compartment disease is considered. These results suggest that malalignment may not be a risk factor for incident OA, but rather, may serve as a marker of disease severity and/or its progression. Further research is needed to confirm these findings. In particular, we would recommend evaluation of the relationship between the mechanical axis and incident knee OA.

Acknowledgements

We would like to thank the study participants and the staff of the Framingham Osteoarthritis Study.

AUTHOR CONTRIBUTIONS

Dr. Hunter had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Hunter, Harvey, Gross, Zhang,

Acquisition of data. Hunter, McCree, Aliabadi, Sack,

Analysis and interpretation of data. Hunter, Niu, Felson, Aliabadi, Zhang,

Manuscript preparation. Hunter, Felson, Harvey, Gross, McCree, Zhang,

Statistical analysis. Hunter, Niu, Zhang.

Ancillary