What is an important difference in gait speed in adults with knee osteoarthritis?

Abstract

Objective.

Little is known regarding what difference in functional performance measures are significant in individuals with chronic medical disease. We examined the important differences in gait speed in adults with radiographic knee osteoarthritis.

Methods.

Functional performance was measured by gait speed using 20-meter and 400-meter walk tests performed at self-selected usual pace among adults with radiographic knee osteoarthritis participating in the Osteoarthritis Initiative at baseline and two years later. Both distribution-based methods and anchor-based methods used to calculate the important differences in gait speed. Anchor-based methods used chair stand rate and self-reported function to estimate gait speed differences related to physical function.

Results.

We included 2527 participants with radiographic knee osteoarthritis. Distribution-based important difference estimates for 20-meter walk ranged from 4.1 to 6.4 meters/minute and 400-meter walk estimates ranged from 2.9 to 6.5 meters/minute. Prevalent (cross-sectional) anchor-based estimates for 20-meter walk ranged from 5.4 to 6.9 meters/minute and for 400-meter walk ranged from 3.0 to 6.9 meters/minute. Longitudinal anchor-based estimates were deemed unreliable. Combining distribution-based with prevalent anchor-based methods showed an important gait speed difference for 20-meter walk is between 4.1 and 6.9 meters/minute and for 400-meter walk is between 2.9 and 6.9 meters/minute.

Conclusion.

Our results found the important difference in gait speed for 20-meter walk and 400-meter walk are consistent with important difference estimates for older adult populations. These findings can provide benchmarks for assessing and understanding functional performance outcomes when comparing exposure groups and can be used in designing future studies targeting adults with radiographic knee osteoarthritis.

Chronic arthritis, including knee osteoarthritis (OA), is a leading cause of functional limitations and disability. [1–3] The ability of older adults to maintain physical function is central to independent living, including persons with knee OA and other chronic diseases. Recognizing the importance of functional ability for overall quality of life, functional performance is frequently assessed in clinical and epidemiology studies. An important issue in the design of clinical studies is to understand what magnitude of difference in the outcome of interest is meaningful. While studies in the general population address this question for a variety of measures [4–7], the smallest differences that are meaningful or important in functional performance measures such as gait speed are not known for people with chronic conditions, particularly individuals with knee osteoarthritis.

The clinically important difference has been defined as “the smallest benefit of value to patients”[8] reflecting the amount of improvement that is important to patients and, assuming minimal side effects and reasonable cost, would change a patient’s management. [9] Methodological approaches to detect important differences can be broadly classified into two categories: distribution-based and anchor-based approaches. Distribution-based approaches rely solely on statistical criteria to distinguish true differences from random error. In contrast, anchor-based approaches compares changes in scores with an “anchor” as reference. Conceptually, an anchor-based approach examines differences in health status relevant to a patient and determines the corresponding difference in the measure of interest to estimate a clinically important difference. It is ideal to synthesize the two methods by first using distribution-based methods to identify a range of values for the important difference and validate using one or more relevant clinical indicators. The objective of this study is to estimate meaningful prevalent (cross-sectional) and longitudinal differences for gait speed measured by both the 20-meter walk and the 400-meter walk in adults with radiographic knee OA using a combination of distribution-based and anchor-based methods.

PATIENTS AND METHODS

Study population:

This study analyzed public data from the Osteoarthritis Initiative (OAI), a multi-center prospective natural history study investigating the development and progression of knee osteoarthritis (OA). OAI study design and eligibility criteria have been described in detail elsewhere.[10] The OAI recruited 4796 men and women between 45 to 79 years of age with or at high risk of developing symptomatic radiographic knee OA. Approval was obtained from the institutional IRB at each participating OAI site and at Northwestern University, Chicago, IL, and written informed consent was obtained from each participating subject. For this analysis, we used OAI public data from the baseline and 2-year follow-up visit.

We examined the 2540 participants with baseline radiographic knee OA defined as Kellgren Lawrence (K/L) grade ≥2. The analysis cohort was limited to 2527 participants who contributed baseline gait speed data (2527 for 20-meter walk gait speed, 2425 for 400-meter walk gait speed). Among these, 2220 contributed 2-year follow-up gait speed data (2220 for 20-meter walk gait speed, 1939 for 400-meter walk gait speed).

Gait speed:

Functional performance was measured by gait speed including 1) 20-meter walk and 2) 400-meter walk. The 20-meter walk is used in many epidemiological studies to measure functional performance and is a standard outcome measure for osteoarthritis. [11, 12] Participants were instructed to walk at their usual walking pace over a 20-meter course in an unobstructed, dedicated corridor. Assistive devices such as a cane were permitted. We calculated the 20-meter gait speed measured in meters/minute. OAI data is the average of the two 20-meter walk trials if a participant finished both trials. The 20-meter walk was performed annually at OAI study visits.

The 400-meter walk assesses physical endurance and fitness in addition to a participant’s ability to walk. Long distance corridor walk has been shown to predict subsequent mobility limitations, disability, cardiovascular disease onset, and mortality in older adults [13]. Participants received instructions to walk 20 laps of a 20-meter course in an unobstructed, dedicated corridor at a maintainable pace. Assistive devices and standing rests were permitted. Gait speed for the 400-meter walk test was calculated in meters/minute. The 400-meter walk was performed biennially at OAI study visits.

Data from incomplete performance tests were not used.

Distribution-based analyses:

Distribution-based estimates rely on the statistical distribution of the outcome of interest (gait speed). We used three distribution-based techniques: 1) standard error of measurement (SEM), 2) 0.5 times the standard deviation (SD) of change [14], and 3) 0.5 times the SD of the baseline measurement. The SEM is a measure of precision of the scale and can be interpreted as the smallest difference or change likely to reflect a true difference or change rather than measurement error. It reflects the minimally detectable difference in a scale. The SEM is defined as ${\text{SD}}_{\text{baseline}}\sqrt{1-\text{ICC}},$ where ${\text{SD}}_{\text{baseline}}$ is the standard deviation of data collected at baseline evaluation, and ICC is intraclass correlation coefficient which measures test-retest reliability [15]. ICC for 20-meter walk was estimated from test-retest in-house baseline data, (ICC=0.90, data not shown). ICC for the 400-meter walk test was obtained from the literature (ICC=0.95)[16]. We calculated the SEM and 0.5 times the SD using baseline data for both gait speeds. We used longitudinal data (SD of change from baseline-to-follow-up) to estimate the SD of change over two years.

Anchor-based analyses:

Anchor-based methods compare an independent measurement to gait speed. Anchors can be cross-sectional, comparing gait speed between clinically distinct groups, or longitudinal, comparing change in gait speed with change in an independent measure. We considered an anchor usable if the sample size was at least 10 for each clinically distinct group, the correlation between the anchor and gait speed of at least 0.3, and the corresponding effect size was within a plausible range of 0.2–0.8.[17] We examined clinically meaningful anchors measured by patient-reported Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function measure, the patient-reported Medical Outcomes Study Short Form 12 Physical Component Summary Scale Score physical function measure (SF-12 function), and chair stand performance rate objective test of lower extremity function as anchor measures examining differences in gait speed between tertiles for each anchor.[18] WOMAC was modified to ask about right and left knee symptoms separately. We calculated person-level scores using the more symptomatic of the 2 knees. WOMAC function scores range from 0 (best function) to 68 (worst function). SF-12 function ranges from 0 (worst function) to 100 (best function).[19] Chair stand performance rate (stands/minute) was calculated from the time required to complete 5 repetitions of rising from a chair and sitting down.[20] These measures were assessed at baseline and 2-year follow-up. In the longitudinal analyses, we calculated the difference in change in gait speed over 2 years among individuals who had a substantive anchor improvement or decline compared to those who did not change substantively for each of WOMAC function, SF-12 function, and chair stand rate. We considered an anchor change to be substantive if it exceeded ½ of the baseline anchor standard deviation. We summarized the gait speed anchor-based important differences estimates by calculating the minimum, maximum, median, 25^th percentile, and 75^th percentile.[17]

All analyses were performed using SAS software version 9.4. A nominal 5% alpha significance level was used in the statistical tests.

RESULTS

Participant characteristics:

As shown in Table 1, the majority of the analytic sample was female (58%) with mean age of 62.6 (SD=9.0). Most participants were either overweight (39%) or obese (45%). Almost half (46%) had severe baseline knee OA (K/L grade = 3 or 4). Mean gait speed was 78.7 (SD=12.9) meters/minute for 20-meter walk and 77.8 (SD=13.0) meters/minute for 400-meter walk. Figure 1 shows the distribution of gait speed at baseline for both 20-meter and 400-meter walk.

Table 1:

Baseline characteristics.

	All participants (n=2527)
Age (years), mean (SD)	62.6 (9.0)
Sex (% female)	57.8%
BMI
<25 kg/m2	16.6%
Overweight: 25–29.9 kg/m2	38.8%
Obese: >30 kg/m2	44.6%
K/L Grade (%)
K/L Grade 2	53.6%
K/L Grade 3	35.0%
K/L Grade 4	11.4%
WOMAC function3, mean (SD)	12.9 (12.7)
SF-12 function4, mean (SD)	47.7 (9.3)
Chair stand rate (stand/sec)5, mean (SD)	0.5 (0.1)
20-meter gait speed (m/min), mean (SD)	78.7 (12.9)
400-meter gait speed (m/min)6, mean (SD)	77.8 (13.0)

^1.BMI: Body mass index, n=2523 participants

^2.K/L Grade: Kellgren and Lawrence Grade, scale ranges from 0–4 with 0 no radiographic knee OA and 4 most severe radiographic knee OA

^3.N=2521 participants. WOMAC physical function: Western Ontario and McMaster Universities Osteoarthritis Index range from 0 (best function) to 68 (most functional limitations)

^4.N=2499 participants. Medical Outcomes Study Short Form 12 Physical Component Summary Scale Score range from 0 (worst function) to 100 (best function)

^5.N=2382 participants

^6.N=2425 participants

Figure 1.

Histograms showing baseline gait speed for 20-meter (n=2527 persons) and 400-meter (n=2425 persons) walk.

3.2. Calculation of the important difference:

Table 2 summarizes distribution-based important different estimates. The important difference ranges from 4.1 to 6.4 meters/minute for 20-meter walk and 2.9 to 6.5 meters/minute for 400-meter walk.

Table 2:

Distribution-based important difference estimates of gait speed

Methods	Important difference calculation	20-m gait speed (m/min)	400-m gait speed (m/min)
		(n=2527)	(n=2425)
SEM *	${\text{SD}}_{\text{baseline}}\sqrt{1-\text{ICC}}$	4.1	2.9
Change SD	0.5×SDΔ	4.3 **	3.9 ***
Baseline SD	0.5× SDbaseline	6.4	6.5

^*Test-retest reliability (intraclass correlation coefficient, ICC) is 0.90 for 20-meter gait speed and 0.95 for 400-meter gait speed.

^**n=2220

^***n=1939

SEM: Standard Error of Measurement

SD_baseline = standard deviation of baseline gait speed

SD_Δ = standard deviation of change in gait speed from baseline to 2 year follow-up evaluation

Anchors, per pre-specified cross-sectional criteria, were each related to baseline gait speed by Pearson correlation coefficients greater than 0.3 with sample size greater than 10 for all anchor-based groups. Table 3 shows cross-sectional anchor-based gait speed differences comparing individuals in the middle tertile to individuals in the best or worst tertile for each anchor and the corresponding effect size for each difference. Per the pre-specified criteria, we only considered anchor-based important difference estimates to be usable if the gait speed differences had a corresponding effect size of 0.2–0.8. Usable anchor-based cross-sectional gait speed important difference estimates range from 3.8 to 6.9 meters/minute for 20-meter walk and 3.0 to 6.9 meters/minute for 400-meter walk. Anchors were also examined for pre-specified longitudinal criteria. For all anchors, the correlation between two-year change in gait speed and two-year change in WOMAC, SF-12 function, and chair stand were all <0.2 thus we did not consider the longitudinal anchor-based important differences to be usable (results not shown).

Table 3.

Cross-sectional anchor-based important difference estimates of gait speed

	Anchor	Mean Difference (m/min)		Effect Size
		Best vs. Middle*	Middle vs. Worst*	Best vs. Middle	Middle vs. Worst
20-m gait speed	WOMAC functiona	2.2	6.9	0.19	0.56
	SF-12 functionb	5.4	6.4	0.48	0.51
	Chair Stand Ratec	3.8	6.6	0.33	0.56
400-m gait speed	WOMAC functiond	3.0	6.2	0.25	0.49
	SF-12 functione	5.7	5.7	0.49	0.45
	Chair Stand Ratef	4.6	6.9	0.39	0.58

^a.n=2521 persons

^b.n=2499 persons

^c.n=2382 persons

^d.n=2420 persons

^e.n=2399 persons

^f.n=2293 persons.

^b.WOMAC function: Western Ontario and McMaster Universities Osteoarthritis Index range from 0 (best function) to 68 (most functional limitations). SF-12 function: Medical Outcomes Study Short Form 12 Physical Component Summary Scale Score range from 0 (worst function) to 100 (best function).

^*Difference in gait speed between individuals in the best tertile and middle tertile for each anchor and for individuals in middle and worst tertile for each anchor.

Bold indicates important difference was considered usable (sample size at least 10, correlation between the anchor and gait speed at least 0.3, and corresponding effect size within a plausible range of 0.2–0.8).

As with any empirically derived value, there is uncertainty and variability associated with important differences. To address this, we provide ranges for important differences rather than single-point estimates. Figure 2 summarizes the usable anchor-based important difference estimates (i.e., corresponding effect size 0.2–0.8) including the minimum, maximum, median, 25^th percentile and 75^th percentile. Combining results from distribution-based and anchor-based approaches, we found the important difference for gait speed is between 4.1 and 6.9 meters/minute for 20-meter walk and between 2.9 and 6.9 meters/minute for 400-meter walk.

Figure 2. Summary of usable* anchor-based important difference estimates.

* An anchor was considered usable if sample size was at least 10, the correlation between the anchor and gait speed was at least 0.3, and the corresponding effect size was within a plausible range of 0.2–0.8

DISCUSSION

The objective of this study was to estimate meaningful prevalent (cross-sectional) and longitudinal differences in functional performance among people with radiographic knee osteoarthritis using 20-meter and 400-meter walk tests. We utilized three different distribution-based techniques providing important difference estimates ranging from 4.1 to 6.4 meters/minute for 20-meter walk and 2.9 to 6.5 meters/minute for 400-meter walk. We also used anchor-based techniques to estimate important differences in gait speed resulting in a range of 3.8 to 6.9 meters/minute for 20-meter walk and 3.0–6.9 for 400-meter walk.

Little is known regarding what magnitude in physical performance differences represents minimal but meaningful change, including among people with arthritis. Estimating meaningful differences is important for designing well powered studies. Several studies investigated small meaningful longitudinal changes of gait speed based on short distance walk tests among older adults.[21–23] But to our knowledge, this issue has not been addressed specifically for persons with rheumatic disease for whom pain and stiffness may interfere with function. Important difference estimates for gait speed based on short walk tests range from approximately 3 to 10 meters/minute in older adult samples. [21–23] Our 20-meter walk important difference results (4.1–6.9 meters/min) in our population of individuals with knee OA were consistent with the important difference findings from these older populations. Two general population studies investigated important differences from long distance walk test outcomes. [21, 22] However those studies evaluated different performance metrics from our 400-meter gait speed outcomes, which prevented direct comparison of important difference estimates.

The few studies in arthritis cohorts investigating the magnitude of meaningful differences in performance did not evaluate important differences. One study using data from patients with hip OA was designed to evaluate performance differences in walk and chair stand tests related to substantial changes in patient reported outcomes (e.g., “a great deal better” or “a very great deal better”)[24], but this study did not provide insight into the magnitude of differences that may be important. A second study among older adults waiting for hip or knee replacement surgery evaluated changes in performance measures deemed beyond measurement error[25], but no criteria were provided to evaluate the clinical or patient-relevance of the estimate. In this study, we utilized both distribution-based and anchor-based analyses to evaluate clinical importance.

Our study has some limitations. First, the OAI is not a probability sample and its participants met many exclusion criteria. This may limit generalizability to studies including a population with a significantly different distribution of baseline gait speed for example a population recruited for a clinical trial with knee pain but not limited to radiographic knee OA. However, OAI participants are recruited from multiple geographic sites using recruitment targets balanced for age and sex groups and represent a broad spectrum of radiographic knee OA. Second, we were not able to calculate longitudinal anchor-based important differences because the correlation between two-year change in gait speed and two-year change in clinical anchors was low making the results unusable. Finally, as our results estimate important differences in prevalent gait speed between groups, individual patients may perceive meaningful clinical benefit at levels that are greater or less than the important differences found in this study; this may be more pronounced for individuals who have baseline gait speed on either extreme (very fast gait or very slow gait including those unable to complete a baseline walk test). Our results only included individuals who completed either the 20-meter of 400-meter baseline walk test and may not be generalizable to groups where a large number of individuals were unable to complete a baseline walk test. Strengths of our study include using established distribution-based and anchor-based methods to estimate a range of clinically important differences and a large natural-history cohort. Although distribution-based approaches are simple to estimate and account for change beyond random variation, they provide no direct information regarding the interpretation or importance of the observed differences. The anchor-based approach provides clinical context in relation to pain and function to estimate important differences that are clinically meaningful.

Estimation of gait speed important differences provides previously unavailable criteria for evaluating patient group differences in functional performance. These important difference estimates will be useful for performing sample size calculations when planning clinical trials to compare differences between groups who received different interventions and will also improve interpretation of results of clinical trials and epidemiologic studies comparing different groups with different exposures. We anticipate that the choice of an important difference will depend on the nature of the study. For example, a low-cost intervention with minimal expected risks would benefit from a smaller important difference of 4–5 meters/minute for 20-meter walk and 3–4 meters/minute for 400-meter walk while an intervention which has potential for serious adverse events or requires significant resources, a larger important difference estimate of 5–7 meters/minute should be utilized for both 20-meter and 400-meter walk. While these estimates are based on meaningful group differences, for determining a meaningful clinical change in an individual we recommend selecting a value closer to 7 than 3 meters/minute.

This is the first report of important differences in gait speed among adults with rheumatic disease who may have pain and stiffness that interferes with physical function including walking. The important difference in gait speed, based on both distribution-based and anchor-based estimates lies between 4.1 and 6.9 meters/minute for 20-meter walk and 2.9 and 6.9 meters/minute for 400-meter walk. Findings from this study provide references for the design of future studies to assess function in arthritis populations.

Significance and Innovation.

• This study provides estimates of the important difference in gait speed among adults with knee osteoarthritis for the 20-meter walk and for the 400-meter walk.

• Prior studies have investigated the important difference of functional performance measures such as gait speed among older adults in the general population. This is the first article to look at important differences in gait speed among adults with rheumatic disease who may have pain and stiffness that interferes with function.

• Findings from this study provide benchmarks for assessing and understanding functional performance outcomes such as gait speed among individuals with radiographic knee osteoarthritis. These importance differences can be used as references to design future studies to assess function in individuals with arthritis who are at elevated risk for decline in function.