Frontal plane knee malalignment may increase progression of knee osteoarthritis (OA) and hasten functional decline. An accurate nonradiographic measure of knee alignment is necessary because the gold standard measure, the long-leg radiograph, is costly and often unavailable. Moreover, nonradiographic measures of knee alignment have not been validated in an obese population, where knee OA is common. The purpose of this study was to develop and assess the concurrent validity and reliability of a nonradiographic measure of frontal plane knee alignment and demonstrate the accuracy of the measure in an obese population.
Fifty-five subjects (41 women, 14 men; mean ± SD age 62.9 ± 10.3 years) with knee OA were examined. A nonradiographic measure (umbilical method) of frontal plane alignment, using the landmarks of the umbilicus, knee, and ankle, was compared with the radiograph gold standard. Statistical significance was accepted at P < 0.05.
Eighty-nine percent of the participants had a body mass index (BMI) placing them in the overweight or obese category (mean ± SD BMI for all subjects 31.3 ± 6.1 kg/m2). Radiographic measures of alignment ranged from 9.1° valgus to 14.3° varus (76% of the participants had varus alignment, 12% had valgus alignment, and 2% had neutral alignment). Umbilical measures ranged from 1° valgus to 21° varus. The umbilical measure was significantly correlated with the radiographic method (r = 0.75, P < 0.001). The error of the umbilical measure was not significantly correlated with the BMI (r = −0.21, P = 0.13).
The umbilical method of assessing frontal plane knee alignment is a valid surrogate for the radiographic gold standard and retains its accuracy in an obese population.
Osteoarthritis (OA) is a common disease affecting older adults (1) and is a leading cause of disability (2). There is increasing interest in understanding the incidence, progression, and management of OA in this large and heterogeneous population. Studies examining the effect of exercise in individuals with knee OA typically have utilized broad inclusion criteria, often based solely on the type of arthritis and/or joint involved (e.g., OA versus rheumatoid arthritis, knee versus hip) (3). However, recent evidence suggests that individual differences, especially structural factors in and around the joint, may be important for researchers to consider when describing the individual with OA and interpreting their response to an intervention.
One such structural variable of the knee, or “local intrinsic factor,” is frontal plane knee alignment. Malalignment of the knee in the frontal plane alters peak load and load distribution across the joint surface, and has been shown to increase the progression of knee OA (4, 5) and hasten functional decline (4). Despite these findings, few studies have considered these factors when describing the sample or interpreting the results of interventions. Because malalignment affects forces at the joint, the response to mechanical interventions such as exercise may be very different between individuals with different mechanical environments of the knee.
The current gold standard for measuring frontal plane alignment requires a standing, long-leg radiograph (6). Other radiographic views that do not extend to the hip and ankle (anatomic axis) are available, but do not represent the true mechanical axis of the knee joint. As a result, these measures have demonstrated varying degrees of correlation to the gold standard long-leg radiograph (r = 0.65–0.88) (6–8). Widespread use of the gold standard long-leg radiograph (and other radiographic measures) in research is problematic due to expense, lack of availability, and exposure to radiation (6, 7). A readily obtainable nonradiographic measure would therefore be valuable to knee OA investigators evaluating frontal plane knee alignment.
There is a high prevalence of obesity in individuals with knee OA (9). Therefore, nonradiographic measures that rely on physical examination of surface anatomy must address the effect of obesity on the accuracy of the measure. The need for measures of frontal plane alignment that are accurate in the presence of obesity is underscored by the strong correlations between obesity and frontal plane (varus) malalignment in people with knee OA (10). Nonradiographic measures of frontal plane alignment have been reported in the literature using inclinometers, calipers, and goniometers (6, 7, 11). Although these measures have shown correlation with the gold standard radiography (r = 0.70–0.80), none has demonstrated that the accuracy of the measure is maintained in the presence of obesity.
There is a critical need to develop a nonradiographic measure of frontal plane alignment that is accurate in obese patients with knee OA. The purpose of this study was to develop and assess the concurrent validity and reliability of a nonradiographic measure of frontal plane knee alignment and demonstrate the accuracy of the measure in an obese population.
PATIENTS AND METHODS
A subset of participants from two studies was evaluated. Both of the studies were conducted in the Department of Physical Therapy Neuromuscular Research and Rehabilitation Laboratory at the University of Missouri. Only baseline data relevant to the measure of frontal plane alignment are reported here.
Fifty-five community-dwelling adults (41 women, 14 men) age ≥50 years were recruited through advertisements in local media, physicians' offices, and Internet sources. Inclusion criteria consisted of 1) self-reported physician-diagnosed knee OA and meeting the American College of Rheumatology criteria for the classification of OA of the knee (12, 13), 2) willingness to exercise regularly and perform all of the testing sessions over a 6-month period, 3) ability to exercise safely and independently at a moderate level of intensity, 4) knee pain within the last year, and 5) a qualifying level of pain or functional deficit demonstrated by the Western Ontario and McMaster Universities Osteoarthritis Index pain or function scales (pain: 1 response of at least moderate or 2 responses of minimal; function: 2 responses of at least moderate or 4 responses of mild). Exclusion criteria consisted of: 1) age <50 years, 2) inability to independently walk or exercise, 3) physical limitations secondary to a condition that is not modifiable with exercise (e.g., active cancer), 4) health problems that might be worsened with exercise, 5) current participation in conditioning exercise, and 6) total joint replacement of the knee (past or scheduled) or total joint replacement of the hip within the last 6 months. Approval of this study was obtained from the University of Missouri Health Sciences Institutional Review Board.
Radiographic measurement of frontal plane alignment.
The radiographic method used in this study has been reported previously in the literature (4). Subjects stood without shoes with tibial tubercles pointing forward with the x-ray beam centered at the knee. Knee alignment in the frontal plane (varus/valgus position) was measured as the angle (in degrees) formed from the intersecting femoral and tibial mechanical axes (Figure 1). The femoral axis was defined as a line drawn from the center of the femoral head to the center of the femoral intercondylar notch. The tibial axis was represented by a line from the center of the ankle talus to the center of the tibial spine. Measurement of varus/valgus at the knee was conducted with Centricity Pictures Archiving Communications System, version 2.0 software (General Electric Healthcare). All of the measurements were completed by a single radiologist. Varus malalignment was recorded as a positive number, neutral alignment was recorded as 0, and valgus alignment was recorded as a negative number. Values were recorded to the nearest tenth of a degree.
Nonradiographic measurement of frontal plane alignment.
Nonradiographic measures of frontal plane alignment were taken by a single investigator (KG) blinded to the radiographic values. The subjects were instructed to take 3 marching steps in place and then maintain the natural foot position obtained after the 3 steps were completed. Subjects then were instructed to adduct their legs until contact between the legs occurred (e.g., secondary to soft tissue around the knee, lower leg, or medial malleoli).
Proximal, middle, and distal landmarks were identified. The distal landmark consisted of a point equidistant between the medial and lateral malleoli and the middle landmark consisted of a point in the center of the knee at the joint line. Both distal and middle landmarks were located with a caliper. The proximal landmark consisted of the umbilicus. The umbilicus was chosen because of its ease of visual identification and consistent location. The consistency in location allowed the umbilicus to be used as a landmark for such things as a guide for abdominal laparoscopy (14) and the identification of the appendix (15).
Measures were obtained using an extendable goniometer (4.5–18 inches; Alimed). Varus, neutral, and valgus alignment were recorded in a similar manner as in the radiograph. Values were recorded to the nearest degree (Figure 2).
On testing day 1, the study staff explained the study and received signed consent. Frontal plane alignment was measured nonradiographically on testing day 1 and repeated within 2 weeks on a subset of participants to assess intrarater reliability. Radiographic examination of frontal plane alignment was completed during this 2-week period.
All of the data were analyzed using SPSS for Windows, release 16.0.1. 2007 (SPSS). Prior to the data analysis, all of the variables were found to satisfy the requirements of normality, linearity, and homoscedasticity necessary for parametric statistics. Descriptive statistics were calculated for measures of alignment (both radiographic and clinical), age, and body mass index (BMI). BMI was calculated as body mass in kg/height in meters squared. Reliability of the nonradiographic measure from the first to the second assessment was conducted on a subset of participants using an intraclass correlation coefficient and a paired-sample t-test. Correlations between nonradiographic and radiographic measures were determined using Pearson's correlation coefficient (r). The correlation between the error of the nonradiographic measure and BMI was calculated to examine the influence of obesity on the accuracy of the measure. Error of the nonradiographic measure was calculated as the absolute value of the difference between the nonradiographic and radiographic measures. To account for the resulting varus shift in frontal plane alignment using the umbilicus as the proximal landmark (instead of the more lateral femoral head used in the radiographic method), a regression equation was developed using the nonradiographic measure as the independent variable and the gold standard radiographic measure as the dependent variable. Statistical significance for all analyses was accepted at P values less than 0.05.
Age, BMI, and nonradiographic and radiographic measures of frontal plane alignment are shown in Table 1. The right knee exhibited a larger range of radiographic malalignment compared with the left knee (23.4° versus 19.5°). Nonradiographic measures of knee alignment using the right leg were therefore chosen for further analysis to generalize to the widest patient population. Radiographic measures of frontal plane alignment ranged from 9.1° valgus to 14.3° varus. Nonradiographic measures ranged from 1° valgus to 21° varus. Modified Kellgren/Lawrence grading (16) of the right knees revealed greater medial than lateral compartment involvement for the presence of osteophytes, joint space narrowing, and sclerosis (Table 2).
Table 1. Descriptive characteristics of the participants (n = 55)*
Values are the mean ± SD unless otherwise indicated. BMI = body mass index.
62.9 ± 10.3
31.3 ± 6.1
BMI range, kg/m2
Right radiographic measure of frontal plane alignment, degrees
3.1 ± 4.8
Right nonradiographic measure of frontal plane alignment, degrees
11.2 ± 4.8
Knee alignment (radiographic), no. (%)
Table 2. Right knee radiographic severity
Joint space narrowing, %
Intrarater reliability of the nonradiographic measurement of knee alignment was conducted on a subset of 46 subjects with an intraclass correlation coefficient (ICC) of 0.85, and a paired t-test between these measures demonstrated no significant difference between the first and second measures (t = −0.93, P = 0.36). Taken together, results from the ICC and paired t-test suggest that our nonradiographic measure demonstrated high intrarater reliability.
There was a significant correlation between the nonradiographic measure of knee alignment and the radiographic measure (r = 0.75, P < 0.001) (Figure 3). Regression analysis defined the relationship between the radiographic and nonradiographic methods by the following equation: predicted radiographic value = 0.746 (nonradiographic measure) − 5.22. Nonradiographic measures resulting in varus scores are recorded as a positive number and valgus measures scores are recorded as a negative number. The coefficient of determination (percentage of variance explained) of the model was R2 = 0.56.
Measurement error and obesity.
There was no significant correlation between the error of the nonradiographic measure and BMI in our sample (n = 55, BMI range 18.7–46.9 kg/m2 [mean ± SD 31.3 ± 6.1]; r = −0.21, P = 0.13). When we evaluated only obese subjects (BMI >30 kg/m2), there was also no significant correlation between measurement error and BMI (n = 30, BMI range 30.1–46.9 kg/m2 [mean ± SD 35.7 ± 4.2]; r = 0.11, P = 0.57).
The nonradiographic measure developed in this study was an accurate assessment of frontal plane knee alignment in patients with knee OA when compared with the gold standard long-leg radiograph. In addition, the use of a consistent and easily located proximal landmark, the umbilicus, appeared to reduce the measurement error associated with obesity when compared with a similar method previously reported using the anterior superior iliac spine (ASIS) as the proximal landmark (11).
Emerging research suggests that adequately describing knee OA patients beyond such basic characteristics as the joint affected and the type of arthritis is critical (17). Both structural factors in and around the joint (local intrinsic factors) as well as exercise, a principal nonpharmaceutical treatment for knee OA (18), alter the mechanical environment of the joint. Because the incidence and progression of knee OA is most likely multifactorial in nature, the effect of a mechanical intervention such as exercise might be mediated by differences in the mechanical environment of the knee.
One local intrinsic factor that accounts for variations in the forces directed at the knee is frontal plane alignment. Total load and distribution of forces are affected by the orientation of the joint in the frontal plane (3). Because of this alteration of force, it may be particularly important to consider these factors when designing or interpreting the success of an exercise intervention. However, few studies reporting the effect of exercise on pain and function in knee OA have considered these factors. This may explain why the effects of exercise on function and pain have shown improvement in some (19–26) but not all (27, 28) studies. Moreover, significant barriers exist in the collection of these data. The gold standard for measuring frontal plane alignment, the long-leg radiograph, is costly, requires specialized equipment and expertise, and is unavailable to many clinicians and researchers. This study has provided a sound nonradiographic tool to help investigators determine the efficacy of treatment programs tailored to differences in local intrinsic factors such as frontal plane alignment.
Obesity is common in knee OA patients, with 80% being either overweight or obese (9). To our knowledge there are currently no nonradiographic measures of frontal plane alignment that have demonstrated utility in an obese population with knee OA. Nonradiographic measures from two previous studies demonstrated acceptable accuracy (r = 0.70–0.80), but one study limited their subject pool to those with a BMI <36 kg/m2 (6) and the second study did not consider the effect of obesity on the accuracy of the measure (7). A nonradiographic measure validated in our laboratory using a subject pool of knee OA patients with a wide BMI range (mean ± SD 32 ± 5.6 kg/m2, range 20–47) also demonstrated acceptable accuracy when compared with the radiographic gold standard (r = 0.74, P < 0.001). However, the error of this measure, which used the ASIS as the proximal landmark, was significantly correlated with BMI (r = 0.37, P = 0.02), calling into question the use of this landmark in the measurement of frontal plane alignment in individuals with high BMI (11).
In the present study, 49 (89%) of the 55 subjects had a BMI that placed them in the overweight or obese category (BMI ≥25 kg/m2), similar to other knee OA populations reported in the literature (9). We used an extendable goniometer and the landmarks of the center of the ankle, the center of the knee, and the umbilicus (umbilical method), and demonstrated acceptable accuracy when compared with the gold standard (r = 0.75, P < 0.001). In contrast to the method reported previously utilizing the ASIS, the center of the knee and the center of the ankle (ASIS method) (11), the error of the present measure using the umbilicus did not appear to be related to BMI. The umbilical method of the present study and the ASIS method reported previously (11) were identical except for the selection of the proximal landmark. The nonsignificant correlation between the error of the umbilical method and BMI suggests that the umbilicus was more reliably indentified in the subject pool compared with the ASIS.
A possible confounder in the use of the umbilicus as the proximal landmark is the tendency for a caudal drop in the umbilicus in the presence of a large abdominal panniculus (14). To our knowledge, there is no method to quantify the vertical position of the umbilicus in a standing weight-bearing position that would not be unduly influenced by anthropometric differences in such variables as overall height or torso length. It has been noted that not all individuals that are obese have an abdominal panniculus (14); therefore, general measures such as abdominal girth may not capture positional changes of the umbilicus that would affect the accuracy of a measure using the umbilicus as a landmark. BMI has been shown, however, to be significantly related to the size of the panniculus (r = 0.54, P < 0.001) (29). Although we cannot comment on the presence of a caudal shift of the umbilicus in our subjects, the measure was shown to be accurate despite 89% of the participants being overweight or obese. In addition, the umbilical measure continued to demonstrate good accuracy in a subgroup of 30 only obese (BMI ≥30 kg/m2) participants (correlation of measurement error and BMI: r = 0.11, P = 0.57).
One limitation to the utility of the umbilical method is the effect of moving the proximal landmark medially from the ball of the femur to the umbilicus. The umbilical femoral axis (a line from the umbilicus to the center of the knee) will therefore be more medially placed than the gold standard radiographic femoral axis (line from the ball of the femur to the center of the knee) (Figure 4). This medial shift of the umbilical femoral axis will therefore cause a subsequent varus overestimation of the frontal plane alignment when compared with the radiographic measure. In this study, the mean value of the umbilical method was 8.1° more varus than the mean value of the radiographic gold standard. Use of the raw umbilical method values would therefore not represent the “true” radiographic angle and would make interpretation of these values problematic. To account for this varus shift, we have developed an equation to convert the umbilical method measurement of frontal plane alignment to a value that approximates the radiographic measure. This can be accomplished by the use of the following regression equation: predicted radiographic value = 0.746 (umbilical method) − 5.22. Using this model, the umbilical method value is denoted as a positive number if varus, and is denoted as a negative number if valgus. The predicted radiographic value should also be interpreted similarly (varus if positive, valgus if negative). This would allow the interpretation of 0 as neutral and increasing values (either positive or negative) as greater degrees of malalignment.
This study was not adequately powered to produce a definitive regression equation that converts a nonradiographic measure to a corresponding radiographic value. Future studies are needed to validate this regression equation on larger numbers of knee OA patients, especially those with a BMI >47 kg/m2. Interrater reliability would also need to be established. Studies may attempt to improve on the accuracy of the umbilical method by including a variable that would reflect the vertical position of the umbilicus, specifically to account for obese individuals with a large abdominal panniculus. The effect of extreme weight fluctuation on the accuracy of the measure has also not been established.
In summary, the present investigation addressed two critical needs in the study of knee OA: 1) to provide an easily obtained, safe, and accurate nonradiographic measure of frontal plane alignment, and 2) to provide a valid nonradiographic measure of frontal plane alignment for obese individuals. We believe the umbilical method of measuring frontal plane knee alignment will allow for the widespread characterization of knee OA patients by this important local intrinsic factor. The development of this measure is one important step in the characterization of local intrinsic factors that may ultimately improve the treatment of knee OA.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Gibson 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 conception and design. Gibson, Sayers, Minor.
Acquisition of data. Gibson, Sayers, Minor.
Analysis and interpretation of data. Gibson, Sayers, Minor.