Ultrasound measures of muscle thickness may be superior to strength testing in adults with knee osteoarthritis: a cross-sectional study

Subjects in this study included 23 adults with unilateral symptomatic knee osteoarthritis, recruited from the outpatient clinic of the primary investigator. Knee OA was diagnosed using American College of Rheumatology (ACR) guidelines [6] based on clinical and radiographic findings. All subjects were screened by telephone for their suitability for enrollment based on ACR guidelines including pain in the knee and at least one of the following: age greater than 50 years, morning stiffness less than 30 min, and joint crepitus. Subjects were excluded from the study if they had any of the following: a prior corticosteroid injection into the knee within 4 weeks prior to enrollment, a prior diagnosis of a neuromuscular condition that affected lower extremity strength, or an alternative rheumatologic diagnosis explaining their knee pain. If subjects met these criteria, they received a weight bearing anterior-posterior and lateral radiograph of both knees. Based on ACR guidelines, the presence of osteophytes on the symptomatic knee was required for radiographic diagnosis of OA. Once subjects met clinical and radiographic inclusion and exclusion criteria, they were entered into the study. Data were collected by trained research assistants in a single in-person visit. The study was approved by the host institution’s IRB and all patients provided written informed consent.

The Western Ontario and McMaster Universities Arthritis Index (WOMAC) was used to assess subjects’ pain, stiffness, and physical functioning. The WOMAC questionnaire is well validated in adults with knee OA and includes 24 questions that measures the three dimensions of pain, disability and joint stiffness.

The Lower Extremity Activity Scale (LEAS) [7] was used to determine the level of daily physical activity in each patient. The LEAS is a self-administered 18-level questionnaire that has been validated in adults with knee OA.

Anthropomorphic measurements, including height and weight, were obtained to calculate joint torques and normalize muscle thickness measurements. Length of the lower limb was measured from the anterior superior iliac spine (ASIS) to the lateral malleolus, and the lower leg was measured from the lateral femoral condyle to the lateral malleolus. All lower limb measurements were performed by a trained research assistant with the subject supine using a flexible tape measure. The ASIS, lateral femoral condyle, and lateral malleolus were identified by palpation. The average of two separate measures was used for the calculating limb length based on previous reports of optimizing validity of this measurement method [8].

Kellgren Lawrence grading of the radiographic degree of osteoarthritis was performed for both knees by the primary investigator.

Muscle groups evaluated with ultrasound imaging included the knee extensor group (quadriceps femoris); hip abductor group (gluteus medius and minimus); hip adductor group (adductor brevis, adductor longus, adductor magnus, and gracilis); and ankle plantarflexor group (gastrocnemius and soleus). Prior to obtaining ultrasound measures on study participants, we developed a standardized protocol for measuring muscle thickness using normal volunteers to ensure maximal interrater reliability. Two evaluators were trained to perform ultrasound scans following the same protocol. For each muscle studied, we used bony landmarks and surface markings to identify a location as close as possible to the mid-portion of the muscle belly. For the quadriceps and hip adductors, a skin mark was placed at half of the distance between the greater trochanter and the lateral condyle of the femur. This line was extended circumferentially across the anterior and medial leg to obtain consistent imaging of the quadriceps and adductors. Next, a mark was placed at 30% from the distal end of a line between the lateral femoral condyle and the lateral malleolus at the ankle. This corresponded to the mid-portion of the gastrocnemius and soleus. A final mark was placed at half the distance from the ASIS to the greater trochanter of the femur, corresponding to the mid-portion of the gluteus medius and minimus.

A Sonosite X-Porte (Bothell, WA) with a curvilinear 5–2 MHz transducer was used to obtain all ultrasound images. Subjects lay supine on an exam table. The transducer was placed perpendicular to the skin/musculature to minimize risk of sampling a muscle obliquely and to ensure repeatability. After the muscle was identified, the examiner slightly retracted the transducer so as to not compress the muscle; the image was considered to be optimized when a thin film of gel was present between the skin and the transducer indicating that no manual compressive forces were distorting the muscle. Once the ultrasound image was optimized, a still image was captured and the muscle thickness was measured with caliper-based tools included in the machine software (Fig. 1). The process was repeated three times for each muscle group and all three measurements were recorded. Once all images were obtained from one lower extremity, the same method was used for imaging of the other.

A Lafayette Model 01165 hand-held dynamometer (Lafayette, IN) was used to measure peak force over a 3 s period, as per settings on the dynamometer. Anatomical markers were used for dynamometer placement to achieve accurate lever arm measurements. When obtaining measurements for the hip abductors, the subject was placed in the supine position, and the dynamometer was placed 5 cm proximal to the lateral malleolus on the lateral side of the lower leg. The subject was cued to abduct the leg against the resisted pressure of the dynamometer. For the adductors, the subject was again supine, and the dynamometer was placed 5 cm proximal to the medial malleolus on the medial aspect of the lower leg, and instructed to adduct the leg against the resisted pressure of the dynamometer. Finally, for the quadriceps, the subject was seated and the dynamometer was placed in the midline at 5 cm proximal to the lateral malleolus. We chose these locations based on prior studies that indicated high levels of reliability and validity [9‐11].

All of our strength tests were isometric “make tests”, such that the subject pushed against the dynamometer while the examiner maintained the dynamometer as steadily as possible. For each test, the subject was allowed to have one warm-up (~ 50% maximum strength) to account for any habituation. The test was repeated three times for each muscle group. Each subject was given a 30 s rest period after each of the tests performed to avoid fatiguing the subject. All tests lasted 3 s as determined by the dynamometer. The settings on the machine itself were set to stop recording with an audible beep after this time period had elapsed. To initiate each test, the subjects were instructed to “go” then the examiner repeated “push, push, push” to signal the patient to push as hard as possible for the remaining 3 s of the test. After the dynamometer beeped, the examiner told the subject to “relax” to signal the end of the test. Maximal force attained during each attempt was recorded.

Based on prior studies regarding the ideal method of reporting strength in knee OA, we calculated joint torque as the product of the force measured by the dynamometer and the distance from the dynamometer to the axis of rotation of the joint [4]. Additionally, because strength varies with body size in adults with and without OA, [12] we calculated strength relative to body mass in kg.

All analysis was performed using Microsoft Excel 15.1 (Redmond, WA) and STATA 14.1 (College Station, TX), with alpha level for hypothesis testing set at 0.05. Torque was calculated at each joint by multiplying the force obtained by dynamometry by the lever arm of the limb. For instance, knee extensor torque was calculated by multiplying the strength of knee extension by the length of the lower leg, and is reported in units of Newton meters (Nm).

Data were evaluated for normality using the Shapiro Wilk test and normal quantile plots. We used simple descriptive statistics to describe our cohort, and paired t-tests to evaluate for any differences in muscle parameters between symptomatic and asymptomatic limbs. Because some of the strength measures were not normally distributed, we used Spearman’s rho to evaluate the correlation between baseline characteristics and muscle measures as well as between functional measures and muscle parameters. We considered r values < 0.3 to represent a weak association, 0.3–0.7 to represent a moderate association, and > 0.7 to represent a strong association [13].

To evaluate the relationship between muscle measures and WOMAC in more detail, we performed a simple linear regression analysis, with the total WOMAC score as the dependent variable, and muscle thickness or torque as the independent variable. To control for possible confounding, we performed a multivariable linear regression analysis using age and gender as covariates. We chose age and gender as possible confounders based on the conceptual model that muscle bulk and strength are correlated with both of these variables. In the multivariable analysis, we assessed how much the regression coefficient associated with the muscle measure changed after adjusting for each potential confounder. If the regression coefficient from the simple linear regression model changed by more than 10%, then the covariate was felt to represent a confounder, and was included in the final regression model [14].

To determine the reliability of measurements for both ultrasound thickness and muscle force, we calculated intra-class correlation coefficients (ICCs) (2,1), using a two-way mixed effects model [15]. ICC (2,1) was used because we were interested in generalizing findings beyond the two raters in the study. An ICC > 0.75 was considered good and ICC > 0.9 was considered excellent [16].

Subject baseline characteristics are shown in Table 1. Subjects included 12 females and 11 males with average age of 63.8 years. The majority of patients had moderate osteoarthritis based on the Kellgren Lawrence scale, with chronic painful symptoms due to OA and median symptom duration of 2 years. No subjects had grade 4 radiographic osteoarthritis. Some patients had radiographic osteoarthritis on the contralateral, asymptomatic knee, though radiographic osteoarthritis grade was less on the asymptomatic side. Symptoms as measured by the WOMAC index were mild to moderate, with a mean total WOMAC score of 25, on a scale from 0 to 96, where higher scores indicate worse symptoms. Functional daily activity as measured by the LEAS had a mean score of 13.1, on a scale of 1–18, where higher scores relate to greater daily functional activity.

Subject muscle characteristics are presented in Table 2. There were no significant differences in normalized measured strength (Nm/kg) between symptomatic and asymptomatic limbs. Similarly, there were no differences in muscle bulk of any of the investigated muscles between symptomatic and asymptomatic limbs.

Intraclass correlation coefficients (ICCs) for ultrasound measurements were excellent for all ultrasound measures. ICC (2,1) was 0.95 for quadriceps, 0.92 for hip adductors, 0.91 for hip abductors, and 0.98 for ankle plantarflexors. ICC(2,1) for torque at the hip adductors was excellent (0.93), but only good at quadriceps (0.83), hip abductors (0.87), and ankle plantarflexors (0.77). Reliability was markedly better for ultrasound measures than torque measures at the quadriceps, hip abductors and ankle plantarflexors.

Female gender was moderately associated with higher pain as measured by the WOMAC pain sub-scale. No other correlations between baseline characteristics and WOMAC or LEAS scales reached statistical significance.

Muscle bulk correlated negatively with pain scores such that greater muscle bulk was associated with lower pain scores (Table 3). This association was significant for the quadriceps and hip adductors but did not reach significance in other muscle groups. Quadriceps thickness was strongly correlated with function, with greater thickness associated with better function. Other muscle groups showed mild to moderate correlation with function, with significance seen in the symptomatic hip adductors. Symptomatic joint stiffness was not found to correlate with any measured muscle thickness. Age and symptom duration were not correlated with muscle thickness in any muscle groups. Males showed higher values for muscle thickness than females for all muscle groups.

Similar to ultrasound-measured bulk, muscle torque generated by all muscle groups was negatively correlated with pain such that lower muscle torque was correlated with worse pain (Table 3). This correlation reached levels of significance for hip abductors, hip adductors, and plantarflexors on both limbs. Importantly, there was no significant correlation found between pain and quadriceps torque. Analyzing correlation with function, muscle torques were negatively correlated with function, with significant correlation seen in the hip abductors, adductors, and plantarflexors, but not quadriceps.

In the simple linear regression analysis, quadriceps thickness was the only ultrasound measure significantly associated with the total WOMAC score. Conversely, dynamometer-measured strength of the quadriceps was not significantly associated with total WOMAC score, while strength of the abductors, adductors, and plantarflexors did show a significant association. When assessing for confounding by age and gender in the multivariable model, age did not change the regression coefficient by more than 10% for any of the strength or muscle thickness measures and was therefore deemed not a confounder. On the other hand, the addition of gender to the model resulted in a change in the regression coefficient by more than 10%, and so was considered a confounder and included in the final regression model. The full results of the multivariable regression analysis are presented in Table 4. In the final model, the unadjusted beta for symptomatic quadriceps thickness normalized to weight was − 67.2. In other words, for every 1 mm/kg increase in quadriceps thickness, the corresponding total WOMAC score decreased by 67.2. To place this in context, we calculated the minimum clinically important difference in WOMAC for this group as a 10% change in the mean WOMAC score, or 2.4 points. Using the above unadjusted beta, for a 70 kg adult, an increase in quadriceps thickness of 2.4 mm would be associated with an improvement of 2.4 on the WOMAC scale.