- Top of page
- SUBJECTS AND METHODS
- AUTHOR CONTRIBUTIONS
Knee osteoarthritis (OA) is a major public health concern worldwide (1) and one of the foremost causes of chronic disability in older adults (2). Preventive care is dependent upon identification of risk factors for development of incident knee OA. Additionally, since pain is the primary symptom that leads to physician visits as well as functional limitations, it is important to clarify the risk factors not only for radiographic, but also for symptomatic knee OA. In order to develop therapies directed at underlying mechanisms for incident knee OA, there has been a long-standing need for longitudinal assessment of risk factors.
There is evidence that muscle dysfunction is involved in the pathogenesis of knee OA (3–6). Because lower extremity musculature is the natural brace for the knee joint, potentially important muscle dysfunction may arise from either quadriceps weakness or relative weakness of the hamstrings in comparison with the quadriceps, usually assessed as the hamstring:quadriceps (H:Q) ratio. An H:Q ratio of ≥0.6 is considered to be normal (7–9). Therefore, evaluation of muscle dysfunction in relation to the knee joint should examine quadriceps strength as well as the balance of muscle strength.
Numerous cross-sectional studies have shown that persons with knee OA have lower knee extensor strength than control participants without knee OA (3–5, 10–12). Although some of these studies indicate that quadriceps muscle weakness may precede knee OA (3, 4), this weakness has largely been attributed to joint pain that may limit muscle use and lead to atrophy. To our knowledge, there has been only 1 report of longitudinal data suggesting a link between relative quadriceps weakness (strength/body weight) and incident OA, and this was found only in women (13). Even then, this finding was of borderline statistical significance, perhaps because the numbers of men and women developing incident OA in this study were small. In this study, women who later developed knee OA were 18% weaker at baseline than those who did not develop knee OA (13).
The longitudinal study (13) suggested an effect of weakness on the development of radiographic knee OA, but since radiographic knee OA is often unaccompanied by pain (14), the public health implications of this finding are uncertain. In addition to assessing the effect of quadriceps strength on incident radiographic knee OA in a longitudinal study, there is also a need to assess whether quadriceps strength alters risk for incident symptomatic knee OA:de novo knee pain or stiffness in the context of radiographic stigmata of OA. Symptomatic knee OA has been the focus of increasing interest because it parallels clinical OA, agrees with the American College of Rheumatology (ACR) criteria for OA (15), is unlike incident radiographic OA, and has clear-cut clinical and public health implications.
While certain risk factors for radiographic knee OA have been characterized, less is known about risk factors for incident symptomatic knee OA. Therefore, the objective of our study was to evaluate the relationship between quadriceps muscle dysfunction (strength and balance with the hamstrings) and incident radiographic tibiofemoral and incident symptomatic whole knee OA.
SUBJECTS AND METHODS
- Top of page
- SUBJECTS AND METHODS
- AUTHOR CONTRIBUTIONS
The Multicenter Osteoarthritis (MOST) Study is a prospective, longitudinal cohort study of risk factors for knee OA. Enrollment was from a volunteer sample of individuals from 2 communities and surrounding regions in Iowa and Alabama. The study enrolled 3,026 men and women (6,052 knees), 50–79 years of age, who responded to mass mailings or advertisements and were screened by telephone for risk factors including age, sex, previous knee injury or surgery, and overweight status based on percentiles derived from the Framingham Heart Study cohort (i.e., women in their 6th, 7th, and 8th decades weighing over 154, 151, and 148 pounds, respectively, and men weighing over 194, 187, and 182 pounds, respectively). Exclusion criteria included a history of (or planned) bilateral knee replacement; cancer with the exception of nonmelanoma skin cancer or breast, cervical, colon, prostate, rectal, or uterine cancer successfully treated with surgery; history of chemotherapy or radiation therapy; rheumatologic disease; or a plan to move out of the area in the next 3 years. Inclusion criteria are outlined in Figure 1.
The study of incident radiographic tibiofemoral OA included 2,519 knees (1,617 participants) without pre-existing radiographic tibiofemoral OA (Kellgren/Lawrence [K/L] grade ≥2) (16) at baseline, who met the inclusion criteria (Figure 1).
Using previous approaches that have characterized symptomatic whole knee OA (17), we excluded knees with frequent symptoms. Knees were considered free from frequent knee symptoms at baseline if participants answered “No” at either the telephone screen or clinic visit in response to the question, “During the past 30 days, have you had any pain, aching, or stiffness in your knee on most days?” The study of incident symptomatic whole knee OA included 3,392 knees (2,078 participants) that did not have symptomatic whole knee OA at baseline and met the inclusion criteria depicted in Figure 1 (16). Below are the evaluations completed at the baseline visit.
Concentric knee extensor strength was assessed with a Cybex 350 computerized isokinetic dynamometer (Avocent, Huntsville, AL) set at at 60 degrees/second and a chair back angle of 85°. HUMAC software,version 4.3.2/Cybex 300 for Windows98 Software Package (CSMI, Stoughton, MA) was used for data acquisition. Participants were provided instructions using a standardized script for subject testing and 3 practice trials using 50% effort. After the practice trials, 4 repetitions were completed for flexor and extensor torque. Participants' concentric knee extensor and flexor strengths (Nm) were considered the peak torque obtained over 4 trials. Trained examiners, certified in the standardized MOST strength testing protocol, underwent annual recertification to assure uniformity in following the strength testing protocol. Examiners calibrated the isokinetic dynamometer position, angular velocity, and torque (at 25 and 245 Nm) monthly.
Participants with unilateral knee replacement performed the test on the contralateral side only. Participants were excluded if they had a systolic blood pressure >199 mm Hg, a diastolic blood pressure >109 mm Hg, history of cerebral aneurysm, cerebral bleeding within the past 6 months, back surgery within the previous 3 months, myocardial infarction or cataract surgery within the previous 6 weeks, untreated inguinal hernia, or pain that precluded strength testing. To avoid potential pain or injury associated with a maximal eccentric contraction, peak torque was recorded concentrically. In a validity study conducted concurrently with the MOST study and using the isokinetic dynamometer, the strength testing protocol had an intraclass correlation coefficient of 0.94 (0.82–0.99), a coefficient of variation of 8% (6–12%), and a within-subject variation of 6.3 Nm (4.71–9.63). For the H:Q ratio, a ratio of the peak torque of the knee flexors to knee extensors was calculated to investigate whether antagonist/agonist imbalance may alter risk for incident OA. The cutoff for dichotomizing the H:Q ratio was defined at 0.6 (7–9). A confirmatory analysis was also performed using 0.8 as a cutoff (8).
Knee radiograph assessments.
Weight-bearing, fixed flexion posteroanterior (18) and lateral radiographs (19) of the knees were obtained at baseline and at 30 months according to the MOST radiograph protocol as previously described (20). Radiographs were taken of the contralateral knee in participants with unilateral knee replacement. Each subject's baseline and followup radiographs were paired and scored by 2 independent readers (an experienced academically-based musculoskeletal radiologist and a rheumatologist experienced in study reading) according to the K/L scale (16). Readers were not blinded to radiograph sequence, but were blinded to subject strength, which was the predictor in this study. For cases where the 2 readers disagreed on the presence of incident radiographic tibiofemoral OA, an adjudication panel of 3 experienced readers decided.
During the telephone screen, trained and certified interviewers asked participants, “During the past 30 days, have you had pain, aching or stiffness in or around your knee on most days?” Knee symptoms were assessed again at the baseline clinic visit, where participants were asked the same question again about knee pain, aching, or stiffness. Participants who responded negatively on either the telephone screen or the baseline visit questionnaire were considered free from knee symptoms at baseline.
At the 30-month telephone screen and clinic visit, participants were again asked the same question regarding pain, aching, or stiffness in each knee on most of the past 30 days. Incident knee symptoms were defined by an affirmative response on both the screen and visit at 30 months.
Femoral neck bone mineral density (BMD).
Because femoral neck BMD has been related to both knee extensor strength and incident OA, we controlled for BMD in our analyses. We obtained a BMD scan of the proximal femur in participants without a history of bilateral hip replacement using dual-energy x-ray absorptiometry (Hologic, 4500a and 4500w, Bedford, MA). BMD of the femoral neck region was recorded in gm/cm2.
At baseline, height (cm; Stadiometer, Holtain, Wales, UK) and weight (kg) were obtained by trained and certified staff, and body mass index (BMI) (kg/m2) was calculated as reported previously (20). Two height measurements were initially taken. If measurements differed by 3 mm, then 2 additional measurements were completed. All measurements were recorded and averaged.
At baseline, participants completed the validated Physical Activity Scale for the Elderly (PASE) questionnaire, and activity scores were calculated (21).
Definition of incident radiographic tibiofemoral OA.
Knees met criteria for incident radiographic tibiofemoral OA if they had no radiographic tibiofemoral OA at baseline (K/L grade <2) and had radiographic tibiofemoral OA (KL grade ≥2) on the 30-month visit radiographs.
Definition of incident symptomatic whole knee OA.
At the baseline and 30-month visits, we obtained radiographs and asked on the phone and at the clinic about the presence of knee pain or stiffness on most days. Incident symptomatic whole knee OA was defined as the combination of knee symptoms and radiographic OA in the tibiofemoral or patellofemoral compartments (whole knee OA) at the followup visit, but not at the baseline visit. Recognizing that OA symptoms fluctuate (22), we thought that at followup participants needed to answer “yes” to the knee symptom questions both times.
Therefore, knees met the criteria for incident symptomatic whole knee (tibiofemoral or patellofemoral) OA if: 1) at baseline they did not have radiographic whole knee OA (x-ray negative) regardless of symptoms, but at the 30-months followup they had the combination of radiographic whole knee OA (x-ray positive) and knee symptoms on both the screen and clinic visit as described above (symptoms positive); or 2) at baseline they had radiographic whole knee OA (x-ray positive) but did not have symptoms on both the screen and clinic visit (symptoms positive/negative or symptoms negative/negative), but at the 30-month followup they had knee symptoms both times when asked (symptoms positive/positive); or 3) they did not have radiographic whole knee OA and symptoms at baseline (x-ray negative or symptoms negative) and underwent knee arthroplasty between baseline and followup as treatment for OA.
Participant characteristics were summarized with frequencies and means. Comparisons of peak strength and the H:Q ratio were made by logistic regression for categorical groups (sex, surgery) and by Pearson's correlation coefficients or linear regression for continuous measures (age, BMI, PASE score). Use of generalized estimating equations (GEEs) is an accepted statistical method for using weighted combinations of observations to extract the appropriate amount of information from correlated data, providing conservative calculations of standard errors in data sets with clusters of correlated data (23). Therefore, we elected to use GEEs to control for between-knee correlations within participants in our knee-based analyses.
We tested the following hypotheses using logistic regression models: 1) high isokinetic knee extensor strength and H:Q ratio at baseline decrease risk for incident radiographic tibiofemoral OA by the 30-month followup visit, and 2) high isokinetic knee extensor strength and H:Q ratio at baseline decrease risk for incident symptomatic whole knee OA at the 30-month followup visit.
Analyses of outcomes were lower extremity based, considering thigh strength and H:Q ratio ipsilateral to each knee. Knees were stratified by sex-specific tertiles of peak knee extensor strength, and whether H:Q ratio was ≥0.6 or <0.6 (7–9). Strata were compared using GEEs, adjusting for the correlation between knees within participants. Known correlates with knee extensor strength or knee OA (age, BMI, hip BMD, history of lower extremity surgery, pain, and PASE score) were included in all multivariable models. Separate analyses were performed for men and women due to differences in strength and the H:Q ratio by sex. SAS, version 9.1 (SAS Institute, Cary, NC), was used for all analyses, and P values less than 0.05 were considered significant.
- Top of page
- SUBJECTS AND METHODS
- AUTHOR CONTRIBUTIONS
To our knowledge, this is the first longitudinal study of a community-based cohort that simultaneously assessed the role of quadriceps muscle strength in the risk for incident radiographic and incident symptomatic whole knee OA. Our results suggest that neither higher knee extensor strength nor normal H:Q balance is protective against the development of incident radiographic tibiofemoral OA. However, for women, being in the highest tertile of knee extensor strength appeared protective against development of incident symptomatic whole knee OA.
Our results concur with the findings of others that strength was lower with the increasing K/L grade of tibiofemoral OA at baseline (11, 12, 24–26) (Table 5). The finding of no correlation between the K/L grade and the H:Q ratio also is in agreement with prior work (11, 27). This may indicate that although quadriceps strength is lower, hamstring strength may also be lower with an increasing K/L grade, resulting in no change in overall sagittal plane muscle balance.
Our longitudinal results appear to differ from those of Slemenda et al, who reported that knee extensor weakness per body weight was a risk factor for radiographic knee OA in women (13). An important difference between our analysis and that of Slemenda and colleagues may account for our respective findings. The prior report divided strength by body weight. Since women with incident OA in that study were more obese, dividing strength by body weight may have resulted in the participants with incident knee OA having a lower strength per body weight due to their increased weight.
The analytic approach used in the prior study recognized that absolute strength values, devoid of context, are of limited usefulness in the assessment of weakness. For example, a certain degree of strength may be either normal or weak depending on body size, age, and sex. However, as fat mass increases with obesity, the ratios of both strength and muscle to body weight will necessarily decrease due to the increased denominator. For example, strength would not be expected to double with a doubling of body mass. Therefore, such 1:1 ratios would not accurately reflect whether the absolute strength should be considered weak.
In order to understand the true relationship between strength and knee OA in people of different body sizes, it is important to assess the relationship recognizing that although a positive correlation between strength and body size is expected (28, 29), like other biologic standards, it may not be a 1:1 ratio (30–33). Therefore, we chose to scale strength by controlling for BMI using logistic regression to determine an index of strength independent of body size, rather than to assume a linear ratio. This difference in methodology may account for the seemingly different results.
If there was a subgroup in which knee extensor strength may be more important for protecting against incident knee OA, we would anticipate that it would have been revealed in a study of the MOST cohort. Participants were recruited to the MOST cohort due to the presence of known risk factors for knee OA, i.e., being overweight or obese, having knee pain, and prior knee surgery or injury. Therefore, the absence of an association between knee extensor strength or balance with incident radiographic tibiofemoral OA in this study of 5,426 at-risk knees with 148 incident radiographic outcomes suggests that a clinically significant association is unlikely to exist.
The MOST study has several unique features that enabled this study to advance knowledge beyond that of prior epidemiologic studies of knee OA, which focused on only radiographic knee OA (13, 34). To our knowledge, this is the first study of the effect of lower extremity strength on the risk for incident symptomatic whole knee OA. The ability to assess risk for development of symptomatic disease allowed this study to provide information most relevant to adults who have activity limitations and present for medical care. Therefore, the results are useful in testing prior supposition that weakness may increase risk for radiographic knee OA (4, 35), and also extend knowledge regarding risk for incident symptomatic whole knee OA using the same cohort and methods. In addition, the MOST study focused on participants representative of those who would most benefit from prevention opportunities, those who have known risk factors for knee OA. Third, this study included comprehensive and reproducible radiographic techniques, as well as measuring the outcome on a cohort of 3,026 participants with very little loss to followup.
A potential limitation is inherent in the measurement of strength. The lower strength in participants with radiographic knee OA at baseline may indicate true weakness, or it may indicate coactivation of knee flexors during knee extensor testing, leading to a lower estimate of knee extensor strength due to simultaneous antagonist action (36). For example, if participants attempted to stabilize their knee during isokinetic testing, using 10 Nm of knee flexor torque during knee extensor testing, this would have reduced the measured knee extensor strength value. Such co-contraction has been reported, particularly in people who report a sensation of knee instability (37). However, since weakness was not predictive of incident knee OA, this potential limitation is unlikely to have influenced the results. Another limitation of the strength assessment was that this study focused on knee extension and flexion strength, and did not include assessments of hip abductor strength. Study of hip abductor strength, which is recognized as important for control of the knee joint (38–40), may be useful in a more comprehensive assessment of risk for incident knee OA.
In this study, we aimed to identify individuals with current knee symptoms. In order to confirm the presence of consistent knee symptoms, individuals were asked about knee symptoms on most of the last 30 days. This question is based on the ACR criteria for establishing presence of symptomatic knee OA (15), and has also been used in the Framingham Knee Study as well as in the Osteoarthritis Initiative. However, since symptoms can fluctuate, this question may not have identified participants who had frequent symptoms at some point outside of the past 30 days. Finally, a study with a greater number of knees, end points, or a longer followup duration may be able to detect a smaller protective effect of knee extensor strength than was possible in this study that followed approximately 5,400 knees over a 30-month period.
The highest tertile of knee extensor strength appears to protect against incident symptomatic whole knee OA. Neither knee extensor strength nor the balance of knee extensors and flexors (H:Q ratio) appear to protect against incident radiographic tibiofemoral OA in either sex. These findings suggest that targeted interventions to reduce risk for symptomatic whole knee OA may be directed toward increasing knee extensor strength, but alternative strategies should be considered for reducing risk for incident radiographic tibiofemoral OA.
- Top of page
- SUBJECTS AND METHODS
- AUTHOR CONTRIBUTIONS
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 submitted for publication. Drs. Segal and Niu had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Segal, Felson, Nevitt.
Acquisition of data. Segal, Torner, Lewis, Nevitt.
Analysis and interpretation of data. Segal, Felson, Niu, Sharma, Nevitt.