- Top of page
- PARTICIPANTS AND METHODS
- AUTHOR CONTRIBUTIONS
Osteoarthritis (OA) of the hip is a major contributor to physical performance limitations in older adults (1). Specifically, hip OA, along with knee OA, affects the ability to walk and climb stairs more than any other diseases (2). Furthermore, compared with individuals with healthy hips, those with hip OA have strength (torque production) impairments of the hip and knee musculature (3, 4) and range of motion (ROM) limitations of the hip joint (5, 6), and these impairments are important correlates of physical performance (5, 7–9). Overwhelmingly, however, these previous studies (5, 7–9) have focused on linear relationships between physical performance and measures of strength and ROM, with minimal or no regard to possible nonlinear associations. The tendency to assume linear associations is so, despite several previous cross-sectional (10–16) studies not limited to individuals with OA that have demonstrated the contrary, i.e., nonlinear relationships may be the norm rather than the exception. Specifically in these previous studies, measures of muscle strength or joint ROM associated nonlinearly with physical performance such that there was a functional threshold below which the strength/ROM measure and performance were closely related, but beyond which the strength/ROM-performance association attenuated substantially.
Understanding whether similar nonlinear associations exist in individuals with hip OA is important. Thus far, current conservative management guidelines have emphasized muscle strengthening for all patients with hip OA (17); however, studies conducted on muscle strengthening exercises in hip OA have reported only modest treatment effects (18–20). Ostensibly, if muscle performance associations were nonlinear, this may, at least in part, account for the modest effects observed. Furthermore, from a public health perspective, knowledge of the functional strength or ROM thresholds would potentially assist clinicians to better identify individuals who are only mildly disabled by hip OA but who are at risk of mobility decline (i.e., preclinical disability). Indeed, to the extent that preventive interventions would be theoretically effective and relatively easy to be implemented in this specific group of individuals, developing a method to identify them is a worthy endeavor.
Based on these considerations, we initiated this cross-sectional study as a first step to examine whether the associations of physical performance with measures of muscle strength and hip joint ROM are nonlinear in individuals with hip OA. Specifically, we quantified 2 muscle-strength measures, knee and hip extensors strength, and 2 hip ROM measures, hip flexion and internal rotation ROM.
- Top of page
- PARTICIPANTS AND METHODS
- AUTHOR CONTRIBUTIONS
The sociodemographic and functional characteristics of the participants are summarized in Table 1. The variance explained by the muscle strength and hip ROM measures along with the respective breakpoints, where applicable, are shown in Table 2. With the exception of hip extensor strength, linear regression revealed no significant associations of residualized gait speed with strength and ROM measures (P ≥ 0.08). For the stair-climb measures, chi-square tests for the difference in deviances between the linear and lowess models were significant (P < 0.01), indicating the possibility of different linear relationships between strength/ROM and stair measures for different ranges of strength/ROM. Accordingly, piecewise regression models were fitted to allow different slopes in 2 different strength/ROM ranges (Figures 2 and 3). Across models, the strength and ROM measures explained approximately 10–20% of unique variance in stair-climb time. With the exception of the association between hip internal rotation ROM and fast paced stair measures, piecewise analyses indicated that all strength/ROM PPT associations were statistically nonsignificant above their respective breakpoints. In contrast, strength/ROM PPT associations were statistically significant below the breakpoints. Finally, all residual plots showed a relatively constant variance of the residuals about the regression lines and, therefore, no visual evidence of heteroscedasticity.
Table 1. Sociodemographic, clinical, and functional characteristics of the study participants (n = 100)*
|Age, years||62 ± 10||45–85|
|Female, no.||60|| |
|Height, meters||1.65 ± 0.09||1.48–1.92|
|Body mass, kg||76.2 ± 15.6||49.5–130.0|
|Body mass index, kg/m2||27.7 ± 4.9||18.8–40.0|
|Waist circumference, cm||93.9 ± 13.8||68.0–130.0|
|Living alone, no.||26|| |
|Comorbidities, no.|| || |
| 0||62|| |
| 1||26|| |
| ≥2||12|| |
|Disease severity, no.|| || |
| K/L grade 2||65|| |
| K/L grade 3||25|| |
| K/L grade 4||10|| |
|Unilateral symptoms, no.||74|| |
|Duration of symptoms, median years||4.8||0.5–40|
|Muscle strength tests (Nm/kg)|| || |
| Knee extensors||1.83 ± 0.59||0.74–3.79|
| Hip extensors||2.1 ± 0.74||0.79–4.17|
|Range of motion tests, degrees|| || |
| Hip flexion||112.7 ± 14.7||70–142|
| Hip internal rotation||25.1 ± 10.7||1–51|
|Gait speed, meters/second||1.2 ± 0.16||0.84–1.60|
|Timed stair test, seconds (n = 99)|| || |
| Self pace||9.7 ± 2.4||5.9–18.4|
| Fast pace||6.9 ± 2.4||3.2–14.9|
Table 2. Breakpoint estimates for hip impairment measures for piecewise regression (n = 100)*
| ||Timed stair test, fast pace (n = 99)||Timed stair test, slow pace (n = 99)||Gait speed, self pace, r2†|
|r2||Breakpoint estimate (95% CI)||No. below breakpoint||r2||Breakpoint estimate (95% CI)||No. below breakpoint|
|Muscle strength tests, Nm/kg|| || || || || || || |
| Knee extensors||0.18||2.5 (1.8–3.3)||88||0.12||2.0 (1.3–2.7)||65||0.02|
| Hip extensors||0.16||2.1 (1.1–3.1)||56||0.14||2.1 (1.4–2.9)||56||0.06‡|
|Range of motion tests, degrees|| || || || || || || |
| Hip flexion||0.20||111 (94–130)||41||0.15||109 (96–121)||36||0.03|
| Hip internal rotation||0.10||NA§||NA§||0.10||26 (8–47)||51||0.00|
Figure 2. Relationship between muscle strength and residualized stair climb performance (positive and negative residualized values represented the extent to which the actual stair time exceeded or fell below what would be predicted from age, sex, pain, and physical activity level, respectively.) A, knee extensors and residualized fast paced timed stair test, B, knee extensors and residualized self paced timed stair test, C, hip extensors and residualized fast paced timed stair test, D, hip extensors and residualized self paced timed stair test.
Download figure to PowerPoint
Figure 3. Relationship between hip range of motion (ROM) measures and residualized stair climb performance (positive and negative residualized values represented the extent to which the actual stair time exceeded or fell below what would be predicted from age, sex, pain, and physical activity level, respectively.) A, hip flexion ROM and residualized fast paced timed stair test, B, hip flexion ROM and residualized self paced timed stair test, C, hip internal rotation ROM and residualized fast paced timed stair test, D, hip internal rotation ROM and residualized self paced timed stair test.
Download figure to PowerPoint
The optimal cut points (using the Youden index) to identify individuals with suboptimal gait speed (≤1.0 meters/second-1) are shown in Table 3. Combining Tables 2 and 3, the piecewise breakpoints and ROC-derived cut points were 1.8–2.5Nm/kg for knee extensors, 1.5–2.1Nm/kg for hip extensors, 25° and 26° for hip internal rotation ROM, and 109° to 115° for hip flexion ROM. Overall, the ROC-derived strength cut points were lower than the breakpoints identified for the residualized stair measures. Table 3 also shows the positive likelihood ratios for the strength and ROM measures, and these estimates ranged from 1.7–4.1. In clinical terms, these ratio values indicate that the odds of suboptimal gait speed increased by 1.5- to 4.1-fold when a given strength/ROM measure falls below the cut point.
Table 3. Optimal cut points for hip measures to identify individuals with functionally inadequate gait speed (≤1.0 meters/second), based on receiver operating characteristic curves (n = 100)*
| ||Cut point (Youden Index)||No. below cut point†||Area under ROC (95% CI)||Sensitivity, % (95% CI)||Specificity, % (95% CI)||Positive LR (95% CI)|
|Muscle tests, Nm/kg|| || || || || || |
| Knee extensors||1.8 (0.39)||36||0.71 (0.58–0.83)||88 (66–97)||51 (40–61)||1.8 (1.4–2.4)|
| Hip extensors||1.5 (0.44)||22||0.76 (0.65–0.88)||59 (36–78)||86 (76–92)||4.1 (2.1–7.9)|
|Range of motion tests, degrees|| || || || || || |
| Hip flexion||115 (0.32)||50||0.68 (0.56–0.80)||76 (53–90)||55 (45–66)||1.7 (1.2–2.5)|
| Hip internal rotation||25 (0.31)||48||0.62 (0.47–0.77)||77 (53–90)||54 (44–64)||1.7 (1.2–2.4)|
- Top of page
- PARTICIPANTS AND METHODS
- AUTHOR CONTRIBUTIONS
Our goal was to examine whether the associations between physical performance and measures of muscle strength and hip ROM departed from linearity in individuals with hip OA. The principal finding was that measures of muscle strength and hip ROM were nonlinearly associated with the residualized stair measures but not with the residualized gait speed measures (Table 3). To our knowledge, these findings have not been previously described in individuals with hip OA.
Within the range of habitual gait speed observed in our study, not only did we fail to observe a nonlinear association between gait speed and measures of strength and ROM, but a linear association was observed only for the hip extensors. Specific to the knee extensors, our results were inconsistent with those of many (11, 14, 31–33), although not all (34, 35) prior studies. However, this disparity in results is probably explained by the difference in performance on the gait speed test. Specifically, the mean habitual gait speed of our participants was 1.2 meters/second, which is the minimum speed required to cross an intersection (36), and therefore was indicative of a generally high functioning sample. Indeed, only 17 participants walked at a suboptimal speed (≤1.0 meters/second), and gait speed varied approximately 2-fold (from 0.84 to 1.60 meters/second) (Table 1). Interestingly, in studies that have observed an association between knee extensors strength and gait measures (11, 14, 31–33), gait speed varied by at least 5-fold; in studies that have found nonsignificant associations (34, 35), gait speed varied by less than 3-fold. Taken together, perhaps our nonsignificant results are a consequence of limited interindividual variation and a corollary to the postulate by Buchner and de Lateur (12); in a relatively high functioning sample such as ours, the association between strength/ROM and gait performance is likely to be attenuated.
Because our range of gait speed was compressed, the fitting of nonlinear functions was not particularly helpful. Consequently, we used ROC analyses to determine cut points that discriminate participants with and without functionally inadequate gait speed. Although only 17 participants walked at a suboptimal speed, the 95% confidence interval (CI) for the AUCs, with the exception of hip internal rotation ROM, exceeded 0.50, indicating that our measures predict significantly better than random. Reviewing the literature, direct comparisons of our cut points cannot be made with previous data because few studies specifically included individuals with OA (13, 29). Still fewer studies examined hip ROM measures (13), and no studies have examined cut points for hip extensors strength or hip internal rotation ROM. Although not directly comparable, our hip flexion ROM cut point is remarkably concordant with the findings of Escalante et al (13), who demonstrated a J-shape association between hip flexion ROM (measured in quartiles) and the odds of walking at a slow speed, in the San Antonio Longitudinal Study of Aging. Specifically, the authors found that the odds increased sharply by 5-fold for persons in the last ROM quartile, and the corresponding quartile cut point was approximately 114°.
Limiting our comparisons to isometric knee extensors measures, our ROC-derived cut point is lower than that of Ploutz-Snyder et al (29), who reported a cut point of 3.0Nm/kg in 103 community-dwelling older adults, and that of Rantanen et al (14), who reported a cut point of 2.3Nm/kg in 773 older adults who had difficulty with activities of daily living. Because Rantanen et al (14) combined strength measures from both extremities, it may be speculated that their cut point was at least indirectly in agreement with ours. With reference to the study by Ploutz-Snyder et al (29), we could not satisfactorily resolve the discordant results. However, using their cut point (3.0Nm/kg) and assuming an average body mass of 70 kg, the average cut point in absolute strength would be 210Nm, a value that exceeds the average strength reported in healthy older adults (37–42). For this reason, the cut point by Ploutz-Snyder and associates (29) appears implausibly high and should, therefore, be interpreted cautiously.
For the residualized stair-climb measures, because the initial exploratory view of the data using lowess smoothers suggested potential nonlinearity, piecewise regression models were constructed accordingly. Before discussing our findings in detail, we should clearly address the key issue relating to our use of piecewise models. Specifically, the piecewise model chosen in this study assumes a sharp transition at the breakpoint. Given the large 95% CIs of the breakpoints (Table 2), it may be argued that strict breakpoints do not exist, and that the transition between the slopes should be gradual. Moreover, it is probable for other factors, e.g., cardiorespiratory fitness (43), standing balance (44), or lumbopelvic (core) control, to modify the strength/ROM-PPT associations, leading to the wide 95% CIs observed. Indeed, as pointed out by the reviewers, physical performance is the result of complex interdependent neuromuscular and physiologic processes, and it is possible for one physiologic system to compensate for another (44). Ostensibly, if impairments in one determinant can be compensated for by other (nonmeasured) determinants of physical performance, the threshold at which muscle strength or hip ROM begins to affect performance will vary across individuals. Thus, it is important to stress that our breakpoints should be interpreted critically and more broadly as the general range of muscle strength or hip ROM values after which the initial linear regression model may not fit the data optimally.
Returning our attention to the residualized stair measures, Figures 2 and 3 suggest mostly nonlinear relationships. However, 2 details of the 2-slope plots merit specific mention. First, less distinct slope changes were observed for the fast paced measures than for the self-paced measures. In fact, for the association between hip internal rotation ROM and the fast paced stair measure, no breakpoint was identified. Second, the strength breakpoints identified for the residualized stair measures were higher than the ROC-derived cut points identified for suboptimal gait speed (Tables 2 and 3). From a mechanistic viewpoint, the latter findings are not surprising. Specifically, our PPTs represent a gradient of increasing movement demand; self-paced gait test is less demanding than self-paced stair test which, in turn, is less demanding than the fast paced stair test. As movement demand increases, fewer participants would possess strength values that were in excess of what was required for the task; hence, strength breakpoints increase concomitantly.
In contrast, interpreting the ROM breakpoints is somewhat difficult and requires speculation. Although our breakpoints were clearly greater than the hip angles observed during stair climbing in hip-healthy adults (45), they may indicate the general ROM necessary for the muscles to operate at an optimal region of the length-tension curve (46). Also, because participants were not allowed to skip steps during the self-paced and fast paced stair tests, this may have resulted in essentially identical breakpoints, at least for the hip-flexion measures.
Our results have clinical implications. Although causality cannot be inferred from cross-sectional data, it is possible that addressing hip ROM deficits that fall below the breakpoints would improve function, which is reminiscent of the substantial functional improvements associated with hip joint mobilizations demonstrated by Hoeskma et al (18) and Currier et al (47). Interestingly, although Hoeskma et al (48) did not identify patient subgroups for whom mobilizations would be most beneficial, the clinical criteria used to diagnose hip OA in their study included less than 115° and 15° for hip flexion and internal rotation, respectively (49). Because these values fell below our breakpoints, one would expect most, if not all, of their patients to benefit from hip joint mobilizations. Equally, Currier et al (47) found that individuals with symptomatic hip OA and hip internal rotation of less than 17° tended to respond favorably to hip mobilizations. Again, because the ROM was below our breakpoint, their results match our expectations. On the other hand, studies conducted thus far on muscle strengthening exercises in hip OA have reported only modest treatment effects (18–20) and did not identify subgroups for whom exercises would be most beneficial. Our data raise the intriguing possibility that effect sizes and study efficiency could potentially be enhanced had patient subgroups been studied. Future studies should explore this possibility.
Besides the limitations of the piecewise models, our study has other limitations. First, our cross-sectional results may be affected by reverse causation, i.e., physical-function limitations could adversely affect muscle strength and hip ROM, and our inferences about the strength/ROM-PPT associations would be clarified if we had longitudinal data and a study population with more variability in gait speed. Second, this study is limited by the absence of consensus that would enable us to define acceptable levels of stair performance, and, therefore, to compare cut points generated from piecewise and ROC analyses. Third and most important, our sample size was small and to produce more precise breakpoints, larger samples are necessary to examine the complex interaction between the strength/ROM measures and other physiologic systems. Accordingly, our study should best be viewed as a hypothesis generating inquiry intended to motivate a larger research agenda.
At the simplest, we have demonstrated that muscle strength and hip ROM measures are associated with physical performance in individuals with hip OA. However, our results extend previous work by suggesting that these associations may be nonlinear. The ability to identify individuals whose physical performance would decline more sharply with every unit decrease in muscle strength or hip ROM would help health care professionals be more successful in targeting interventions. In that context, although the breakpoints in this study should be considered more speculative than definitive, our approach is, nevertheless, a means by which patient subgroups may be identified, and deserves greater attention in future research.