Considerations for the Selection of Time-Limited Walk Tests Poststroke: A Systematic Review of Test Protocols and Measurement Properties

Abstract

Background and Purpose: 

Systematic reviews of research evidence describing the quality and methods for administering standardized outcome measures are essential to developing recommendations for their clinical application. The purpose of this systematic review was to synthesize the research literature describing test protocols and measurement properties of time-limited walk tests in people poststroke.

Methods: 

Following an electronic search of 7 bibliographic data-bases, 2 authors independently screened titles and abstracts. One author identified eligible articles, and performed quality appraisal and data extraction.

Results: 

Of 12 180 records identified, 43 articles were included. Among 5 walk tests described, the 6-minute walk test (6MWT) was most frequently evaluated (n = 36). Only 5 articles included participants in the acute phase (<1 month) poststroke. Within tests, protocols varied. Walkway length and walking aid, but not turning direction, influenced 6MWT performance. Intraclass correlation coefficients for reliability were 0.68 to 0.71 (12MWT) and 0.80 to 1.00 (2-, 3-, 5- and 6MWT). Minimal detectable change values at the 90% confidence level were 11.4 m (2MWT), 24.4 m (5MWT), and 27.7 to 52.1 m (6MWT; n = 6). Moderate-to-strong correlations (≥0.5) between 6MWT distance and balance, motor function, walking speed, mobility, and stair capacity were consistently observed (n = 33). Moderate-to-strong correlations between 5MWT performance and walking speed/independence (n = 1), and between 12MWT performance and balance, motor function, and walking speed (n = 1) were reported.

Discussion and Conclusions: 

Strong evidence of the reliability and construct validity of using the 6MWT poststroke exists; studies in the acute phase are lacking. Because protocol variations influence performance, a standardized 6MWT protocol poststroke for use across the care continuum is needed.

Video Abstract available for more insights from the authors (see Supplemental Digital Content 1, https://links.lww.com/JNPT/A150).

INTRODUCTION

Rehabilitation professionals have recently been encouraged to use time-limited walk tests, such as the 6-minute walk test (6MWT), to evaluate functional walking capacity across the care continuum poststroke owing to the quality of the tests' measurement properties and perceived utility.^1–3 Use of standardized assessments as a cornerstone of evidence-informed physical therapy practice has been well described.^2,4 The existence of assessment recommendations, however, does not ensure their implementation in clinical practice. Based on self-report data, an estimated 11% of physical therapists use the 6MWT and 26% use the 2-minute walk test (2MWT) to evaluate walking poststroke.⁵ Physical therapists who have the primary responsibility for the rehabilitation of walking poststroke⁶ have described how knowledge of psychometric properties, and, more importantly, perceived clinical relevance of these tests influences their decisions to adopt them into routine clinical practice.^1,7,8 To date, research evidence of the speed and distance requirements for community ambulation⁹ and walk test normative values¹⁰ has been synthesized to improve the interpretability of walk test results. Evidence of reliability, construct validity, minimal detectable change, and minimal clinically important difference is needed to provide convincing evidence of the quality and clinical relevance of these tests.

The knowledge-to-action (KTA) framework is a conceptual model of knowledge translation developed by Graham and colleagues¹¹ to guide thinking about the process of knowledge creation and application. The knowledge creation funnel in the KTA model is used to illustrate the filtering process required to develop knowledge products or tools tailored to the user. First-generation knowledge at the base of the funnel refers to the numerous individual sources of information (eg, research articles, and reports) on a topic that are of variable quality and time-consuming to acquire. Second-generation knowledge, or knowledge synthesis, is described as an essential precursor to the development of user-friendly knowledge tools, such as evidence-based algorithms, guides, and guidelines, that will help persuade and prepare physical therapists to adopt time-limited walk tests into clinical practice. Based on the KTA process, synthesis and critical appraisal of the literature is necessary to inform the development of a guideline on the use of time-limited walk tests in people poststroke.

The objective of the study was to appraise and synthesize the research literature describing (1) reliability, measurement error, construct validity, and sensitivity to change; (2) walk test protocol; and (3) the effect of walk test protocol elements (eg, walkway length and encouragement) on test performance for time-limited walk tests in adults poststroke. In addition, our review aimed to identify gaps in the evaluation of measurement properties of time-limited walk tests, and to identify considerations for the administration and interpretation of performance on time-limited walk tests poststroke to enhance acceptance, utility, and value for practicing clinicians.

METHODS

Overview

A systematic review was conducted according to a review protocol that was developed by the research team. The PRISMA checklist was used to guide reporting.¹²

Search Strategy

We searched 7 electronic databases (MEDLINE (Ovid), EMBASE, PubMed, CINAHL, Scopus, PEDro, and The Cochrane Library) from 1946 to July 2013. Search strategies for each database were developed with input from an information specialist. Search terms included cerebrovascular accident, stroke, 6-minute walk test, and a wide variety of terms associated with walk tests (see PubMed search strategy in the Appendix). No limitations were applied during the search. The principal investigator's personal collection of research literature and reference lists of included studies were reviewed for potentially relevant articles.

Selection Criteria

Studies were considered eligible if (1) participants were adults 18 years or older poststroke; (2) the study reported on reliability, measurement error, construct validity, and sensitivity to change, or the effect of walkway length, encouragement, walking aids, and practice trials on performance of time-limited walk tests (for construct validity, studies reporting associations between walk test performance and other variables, regardless of whether this was framed as validity testing, were included); (3) test duration and track distance were reported to enable test replication; and (4) the report was written in English, French, or Spanish. Because the literature was extensive, we only included studies reporting unadjusted correlations and associated P-values or confidence intervals (CIs) between walk test performance and measures of motor function, aerobic capacity, balance, balance self-efficacy, strength, walking, stairs, mobility, physical activity, participation, health-related quality of life, and discharge destination for construct validity; and studies reporting minimal detectable change (MDC), standard error of measurement (SEM), and SEM% for measurement error. We excluded studies in which (1) less than 80% of the study sample comprised people poststroke; (2) the walk test was completed on a treadmill or embedded within another test; (3) the study objective was to validate the comparator measure or where studies assessed the correlation of change scores (for construct validity); and (4) the study was a conference proceeding, dissertation, case report/series, or limited to abstract form.

Study Selection and Data Extraction

Two authors independently screened titles and abstracts. A single author (NMS or PT) determined the inclusion of potentially relevant studies, and extracted data on general study information, study and participant characteristics, walk test protocol, and results from included studies. To ensure data accuracy, a single author (PT) randomly selected and verified data extracted from 30% of included studies.

Method of Quality Assessment

The methodological quality of included studies was assessed using the COnsensus-based Standards for the selection of health Measurements INstruments (COSMIN) critical appraisal tool.¹³ The first version of COSMIN comprises checklists that each relate to a specific measurement property (reliability, measurement error, hypothesis testing) or the interpretability and generalizability of studying findings. Operational definitions were developed to optimize scoring consistency. For example, for reliability and measurement error, we defined a retest time interval (item 6) over which patient stability would be assumed for 3 recovery phases poststroke as 1 day or less (acute), 5 days or less (subacute), and 3 weeks or less (chronic) based on results from longitudinal studies of walking^14,15 and consensus among authors. Adequate sample sizes were defined as 25 or more and 30 or more for assessing reliability/measurement error and construct validity, respectively, based on minimum sample sizes required to reproduce reliability and validity estimates from larger datasets.¹⁶ Type, side, severity, and time poststroke were identified as disease characteristics relevant to the generalizability of study findings. For stroke severity, reporting of scores on a measure of stroke severity, motor function, or functional burden (ie, Functional Independence Measure¹⁷) was considered acceptable. Irrelevant checklist items were removed. The research team developed a checklist assessing sensitivity to change based on the format of the COSMIN checklists.¹³ Checklist response options were “yes/no” for 32 items and 27 items had a third response option of “can't tell.” A single author (NMS or PT) assessed the methodological quality of included studies.

During the review, COSMIN developers published a new scoring system that classified each measurement property as excellent, good, fair, or poor based on the lowest score reported on the corresponding checklist.¹⁸ Thus, a single author additionally applied the new rating system to studies examining reliability and measurement error. Throughout the review, an author not involved in the quality assessment was consulted to resolve uncertainty.

Data Synthesis and Analysis

The intraclass correlation coefficients (ICCs) used to evaluate reliability were interpreted as excellent (ICC ≥ 0.75), acceptable (ICC >0.40 to <0.75) or poor (ICC ≤ 0.40).¹⁹ To enable comparison, SEM and minimal detectable change at the 90% confidence level (MDC₉₀) were computed across studies reporting test-retest reliability estimates and standard deviation of baseline score using the following equations: 1 ²⁰ and .²⁰ For construct validity, constructs were classified using the International Classification of Functioning, Disability and Health.²¹ Correlation coefficients were interpreted as strong (≥0.70), moderate (0.50-0.69), weak (0.30-0.49), or negligible (<0.30).²² Simple correlations were derived from R² values where reported.

Time poststroke was determined by the range/interquartile range or by the mean/median values if the range was not reported. Study participants were considered to be in the acute, subacute, or chronic phase poststroke if they were less than 1 month, 1 to 6 months, or more than 6 months poststroke, respectively. To enable comparison across studies, results were converted to a common metric unit, frequency data were converted to percentages, and values were rounded to a consistent decimal place.

RESULTS

Study Selection

The search and article selection results are presented in Figure 1. Of the 12 180 records identified, 43 articles^23–65 representing 42 studies were eligible and included in the review. Of the 13 authors contacted to clarify and/or obtain select data not reported in published studies, 10 authors (77%) responded with requested information.

Figure 1.: Process of study selection (PRISMA).

Study Characteristics

All included articles were written in English. Five time-limited walk tests, including the 2-, 3-, 5-, 6-, and 12-minute walk test (2MWT, 3MWT, 5MWT, 6MWT, and 12MWT, respectively) were identified. The 6MWT was most commonly evaluated (n = 36, 82%). Table 1 provides a cross-tabulation of the number of evaluations of each measurement property, and the effect of turning direction, walkway length, practice trials, and walking aids on test performance by walk test.

Table 1. - Frequency of Evaluations of Measurement Properties by Walk Test

Measurement Property or Protocol Element Examined	Number of Evaluations
	2MWT	3MWT	5MWT	6MWT	12MWT	Total
Cross-sectional construct validity	–	1	2	32	1	36
Reliability	2	1	1	6	1	11
Measurement error	1	–	1	4	–	6
Predictive validity	–	–	–	3	–	3
Sensitivity to change	–	–	–	–	1	1
Effect of turning direction	–	–	–	1	–	1
Effect of walkway length	–	–	–	1	–	1
Effect of practice trials	–	–	–	1	–	1
Effect of walking aids	–	–	–	1	–	1
Number of articles	2a	1a	3a	36a	2a

Abbreviations: 2MWT, 2-minute walk test; 3MWT, 3-minute walk test; 5MWT, 5-minute walk test; 6MWT, 6-minute walk test; 12MWT, 12-minute walk test.

^aSelect articles reported on more than one measurement property and more than one test.

Appraisal of Study Methodology

Figures 2, 3, and 4 present the item-level COSMIN scores for articles assessing reliability, measurement error, and construct validity, respectively. Across evaluations of reliability and measurement error, common methodological issues were uncertainty of whether administrations were independent (ie, performance on the first test was not provided to the evaluator or participant at retest) (reliability [R] 100%; measurement error [ME] 100%), inadequate sample size (R 73%; ME 67%), and uncertainty whether participants were stable in the interim period (R 55%; ME 50%). The 3 most common methodological issues among evaluations of construct validity were failure to report a hypothesis (83%), inadequate sample size (67%), and failure to report the measurement properties of comparator instruments (50%). Important flaws were reported across all included studies with respect to insufficient description of the walk test protocol, thus limiting replication. For the generalizability checklist, the common methodological issues were failure to report the method used to select participants (60%), important disease characteristics, and/or description of treatment (47%).

Figure 2.: Quality appraisal (COSMIN) for studies examining reliability (n = 11). Gen, generalizability checklist; Rel, reliability checklist; +, yes; −, no; ?, cannot tell. Important flaws across included studies commonly related to insufficient description of the walk test protocol.

Figure 3.: Quality appraisal (COSMIN) for studies examining measurement error (n = 6). Gen, generalizability checklist; ME, measurement error checklist; +, yes; −, no; ?, cannot tell. Important flaws across included studies commonly related to insufficient description of the walk test protocol.

Figure 4.: Quality appraisal (COSMIN) for studies examining construct validity (n = 36). Gen, generalizability checklist; HT, hypothesis testing; +, yes; −, no; ?, cannot tell. Important flaws across included studies commonly related to insufficient description of the walk test protocol.

Participant Characteristics and Walk Test Protocols

The table (Supplemental Digital Content 2, https://links.lww.com/JNPT/A151) provides details of participant characteristics across articles. The number of articles describing people in the acute, subacute and chronic phase poststroke was 2 (5%), 4 (9%), and 31 (72%), respectively. In 6 studies, the sample consisted of a mix of participants in the acute and subacute (7%, n = 3) and subacute and chronic phases poststroke (7%, n = 3). Usual assistive devices used were reported across 32 (74%) articles.

This table (Supplemental Digital Content 3, https://links.lww.com/JNPT/A152) describes the walk test protocols used in the 42 studies described in 43 articles. All studies reporting the 2MWT, 3MWT, and 5MWT evaluated walking using a straight walkway with test distances ranging from 5 to 30 m.^23–28 The 12MWT was examined using a rectangular walkway 42 m⁶³ and 122 m⁶² long. Of the 35 studies describing 36 protocols for the 6MWT, 76% (n = 28) of the walk tests were performed along a straight walkway ranging from 10 to 85 m in length. The 3 most frequently reported straight walkway distances were 30 m (36%), 30.5 m (11%), and 33 m (11%). Rectangular,^30,32,38,63 oval,^41,49 and circular⁴⁴ walkways were also used. Of the 36 6MWT protocols reported, 14% were the American Thoracic Society 6MWT protocol.⁶⁶

Across the 44 walk test protocols, participants were instructed to walk at either a comfortable (n = 13) or fast (n = 4) pace where pace was described. The position of the evaluator was reported in 13 walk test protocols (30%). Evaluators walked behind (n = 10),^{23,24,30,35,42,45,53,56,59,65} or beside (n = 2)^27,41 participants, or remained at the starting line (n = 1).⁵⁵ Authors described that the evaluator either guarded the participant for safety concerns during the walk test⁴⁸ or provided assistance for balance, weight-shifting, or leg advancement as necessary.^41,56,62

Influence of Walk Test Protocol Elements on Test Performance

Effect of Turning Direction and Walkway Distance

In one study,⁵³ 26 people with chronic stroke walked significantly further and took significantly fewer turns, on average, as walkway length increased from 10, to 20 to 30 m. For example, walk distance improved by an average of 38.2 and 40.6 m after increasing the walking length from 10 to 30 m, respectively.⁵³ Turning direction (to the paretic/nonparetic side) did not influence 6MWT performance.⁵³

Effect of a Practice Trial

In a study⁴² of 83 people within the first-year poststroke, the mean difference ± standard deviation in 6MWT performance across 2 trials performed approximately 30 minutes apart was 0 ± 35 m. Across trials, 58% improved their performance (median 13 m), 40% deteriorated (median −17 m), and 2% did not change.

Effect of Walking Aids

Twenty-five people undergoing inpatient rehabilitation poststroke completed 3 6MWTs on 3 consecutive days using 3 walking aids (4-point cane, simple cane with ergonomic handgrip, Nordic stick) in random order.⁴⁴ On average, participants walked significantly further when using the simple cane (115.5 ± 55.0 m), compared with the 4-point cane (101.4 ± 54.1 m) and the Nordic stick (98.0 ± 51.3 m).

Reliability and Measurement Error

Across the 11 articles reporting reliability, interrater, intrarater, and test-retest reliability was reported in 1, 2, and 9 articles, respectively (Table 2). Test-retest reliability and measurement error were examined in participants in the subacute (n = 2), subacute/chronic (n = 1), and chronic (n = 6; note: one study did not report measurement error) phases poststroke. Inter- and intrarater reliability was examined in a group of participants in the acute/subacute (n = 2) phase poststroke. Across walk tests, point estimates for ICCs and lower 95% CI limits were 0.90 or greater (range 0.90-1.00) with 2 exceptions. The ICC for test-retest reliability of the 6MWT in people requiring physical assistance to walk was 0.80 (95% CI 0.44-0.94),⁴¹ and the ICC for intra- and interrater reliability of the 12MWT was 0.71 and 0.68, respectively.⁶²Table 2 presents the SEM and computed or reported MDC values for the 2MWT (n = 1), 5MWT, (n = 1) and 6MWT (n = 6).

Table 2. - Reliability and Measurement Error (11 Studies)

Author	Walk Test	Level of Gait Deficit	Walk Test Protocol				Re-test Time Interval	N	Test-Retest Reliabilitya ICC (95% CI)	Measurement Error, m	COSMIN Quality Score
			Path Distance, Shape, Location	#Practice/Test Trials, Position of Rater, Timing Tool	Walk Aids Allowed	Encouragement (Timing), Assistance Allowed
Acute (<1 mo) and subacute (1-6 mo)
English et al23	2MWT	NR	10 m, straight, indoor	0/1b, behind, stopwatch	Yesb	Yesb (NR), NR	4 wk, video-taped trials rescored	10	Intra-R: 1.00 (NR)		Poor
Kosak and Smith62	12MWT	FIM walking subscore: 3 ± 1	122 m, rectangle, indoor	NR, NR, NR	Yes	NR, yes	1 d	18	Inter-R: 0.68 (P < 0.0007)		Poor
							Day 1, 3	18	Intra-R: 0.71 (P < 0.0003)
Subacute (1-6 mo)
Da Cunha-Filho et al27	5MWT	NR	5 m, straight, indoor	1/1; behindb, stopwatch	Yesb	Nob, NR	≥10 min	9	0.97 (NR)	SEM = 6.9 MDC95 = 19.2 MDC90 = 24.4c	Poor
Fulk et al41	6MWT	NR	46/76 m, oval, indoor	0/1b, beside, NR	Yesb	No, yes	1-3 d	37 24, walk on owng 13, walk with assistg	0.97 (0.92-0.99) 0.97 (0.91-0.99) 0.80 (0.44-0.94)	SEM = 23.2 MDC90 = 54.1 MDC90 = 52.1c SEM = 26.1 MDC90 = 61.0 MDC90 = 59.6c SEM = 16.7 MDC90 = 39.0 MDC90 = 33.9c	Fair
Subacute (1-6 mo) and chronic (>6 mo)
Liu et al42	6MWT	CGS ≥0.7, 0.3-0.7, <0.3 m/s: 38%, 42%, 21%	20 m, straight, indoor	0/1, slightly behindb, stopwatchb	Yes	Yes (30 s), nob,d	30 min	83	0.98 (0.97-0.99)	MDC90 = 39.2c	Fair
Chronic (>6 mo)
Hiengkaew et al24	2MWT	NR	20 m, straight, indoor	0/1b, behindb, stopwatch	Yes	Nob, NRd	5-10 d	61 12, MAS = 0e 32, MAS = 1-1+e 17, MAS ≥2e	0.98 (0.97-0.99) 0.98 (0.92-0.99) 0.98 (0.96-0.99) 0.96 (0.90-0.99)	SEM = 4.8 MDC90 = 11.4c SEM = 5.1 MDC90 = 11.5c SEM = 4.8 MDC90 = 11.2c SEM = 4.7 MDC90 = 11.6c	Fair
Sakai et al25	3MWT	NR	30 m, straight, indoor	NR, NR, NR	Yes	NR, NRd	2-5 d	14	0.90 (NR)		Poor
Eng et al30	6MWT	CGS: 1.0 ± 0.3 m/s	42 m, rectangle, indoor	0/1b, behindb, stopwatchb	Yes	Yesb (as per ATS), nob	Additional days	12	0.99 (NR)	SEM = 12.4 MDC90 = 28.6c	Poor
Ng and Hui-Chan65	6MWT	NR	33 m, straight, indoor	0/1b, behindb, NR	Yesb	Yesb (as per ATS), NRd	Different days in 1 wk, same time	10	0.98 (0.94-0.99)	MDC90 = 29.0c	Poor
Flansbjer et al31	6MWT	CGS: 0.9 ± 0.3 m/s	30 m, straight, indoor	0/1b, at mid-walkwayb, stopwatch	Yes	No (informed when 3 min left), nob	7 d, same timef	50	0.99 (0.98-0.99)	SEM = 18.6 SEM% = 4.8 MDC90 = 43.3 MDC90 = 30.7c	Fair
Wevers et al56	6MWT	FAC score of 3/4/5, n: 1/3/23	30 m, straight, outdoors	0/1b, behind, GPS/MW	Yesb	Yes (1 min), yesd	1.5 ± 1.3 d (maximum 5 d)	27, GPS	0.96 (0.96-0.98)	SEM = 18.1 MDC95 = 50.2 MDC90 = 42.1c	Poor
Zalewski and Dvorak57								27, MW	0.98 (0.98-1.00)	SEM = 11.9 MDC95 = 33.0 MDC90 = 27.7c

Abbreviations: A, acute (<1 month); ATS, American Thoracic Society; C, chronic (≥ 6 months); CI, confidence interval; CGS, comfortable gait speed; GPS, global positioning system; FAC, Functional Ambulation Classification; FIM, Functional Independence Measure; ICC, intraclass correlation coefficient; Inter-R, interrater reliability; Intra-R, intrarater; MAS, Motor Assessment Scale; MDC₉₀/MDC₉₅, minimal detectable change at 90%/95% confidence level; MW, measuring wheel; NR, not reported; SA, subacute (1-6 months); SEM, standard error of measurement; 2MWT, 2-minute walk test; 3MWT, 3-minute walk test; 5MWT, 5-minute walk test; 6MWT, 6-minute walk test; 12MWT, 12-minute walk test.

^aTest-retest reliability presented unless otherwise indicated.

^bData obtained from author.

^cComputed by authors using published data.

^dParticipants were required to walk independently to be eligible.

^eParticipants categorized based on Modified Ashworth Scale score for ankle plantarflexor tone, where 0 = no increase in tone; 1-1+ = slight increase in tone; ≥2 = marked/considerable increase in tone or showing rigid ankle plantarflexors.

^fFor 2 participants, the interval was 10 and 13 days; for 3 participants, the test session was not performed at the same time of day.

^gParticipants categorized based on FIM locomotion score, where ≥5 = walk without assistance; <5 = required physical assistance to walk.

Construct Validity

Table 3 presents 89 correlation coefficients for relationships between measures of targeted constructs and performance on the 3MWT (1 correlation), 5MWT (3 correlations), 6MWT (81 correlations), and 12MWT (4 correlations). Across the acuity levels, 67% (n = 60) of the correlation coefficients were evaluated among participants in the chronic phase poststroke. No studies examined relationships with discharge destination. Of the 89 correlations, 6 were predictive in nature examining the ability of the 6MWT to predict physical activity^45,49 and health-related quality of life.³⁵Table 4 summarizes the reliability, measurement error, and construct validity findings by walk test and recovery phase poststroke.

Table 3. - Construct Validity (N = 36)

Walk Test	ICF Classification	Construct	Measure	Results Pearson r (P-value, n) Spearman ρ (P-value, n)
Acute (<1 mo)
5MWT	Activity	Walking independence	Functional ambulation classification	r = 0.55 (P < 0.05, n = 20)26
		Walking speed	5-m fast walk test	r = 0.80 (P < 0.01, n = 20)26
Acute (<1 mo) and subacute (1-6 mo)
6MWT	Body function	Aerobic capacity	o 2 peak/age-predicted o 2 max	r = 0.84 (P < 0.001, n = 17)29
	Activity	Mobility	Timed Up and Go	ρ = −0.73 (P < 0.001, n = 41, discharge)34
				ρ = −0.80 (P < 0.001, n = 41, admission)34
		Walking speed	10-m comfortable walk test	r = 0.91 (P < 0.05, n = 17)29
			10-m fast walk test	r = 0.89 (P < 0.05, n = 17)29
Subacute (1-6 mo)
6MWT	Body function	Aerobic capacity	Maximal exercise test duration	r = 0.60 (P < 0.001, n = 36)36
	Activity	Walking speed	5-m comfortable walk test	r = 0.79 (P < 0.001, n = 34)36
				r = 0.89 (P < 0.000, n = 37)41
			5-m fast walk test	r = 0.82 (P < 0.001, n = 30)36
Subacute (1-6 mo) and chronic (>6 mo)
6MWT	Body function	Aerobic capacity	Peak oxygen consumption (o 2 peak)	ρ = 0.34 (NS, n = 30, indoor 6MWT)40
				ρ = 0.39 (P ≤ 0.05, n = 30, outdoor 6MWT)40
				r = 0.39 (P > 0.05, n = 34)30
				r = 0.40 (P < 0.005, n = 63)32
				r = 0.45 (P < 0.05, n = 48)55
				r = 0.56 (P < 0.001, n = 33)36
				r = 0.58a (P < 0.05, n = 48)55
				r = 0.64 (P < 0.001, n = 74)39
				r = 0.68 (P < 0.05, n = 14)52
				r = 0.71 (p = 0.04, n = 8)54
				r = 0.73 (p = 0.027, n = 9)48
	Body function	Balance self-efficacy	Activities-specific balance confidence scale	ρ = 0.43 (95% CI 0.24-0.59, n = 86)35
	Activity	Walking speed	10-m comfortable walk test	r = −0.74 (correlation with time, P ≤ 0.01, n = 21)45
				r = 0.84 (P < 0.001, n = 50)31
	Activity	Walking speed	10-m fast walk test	r = 0.94 (P < 0.001, n = 50)31
	Non-ICF	Health-related quality of life	SF-36-PF Scale	r = 0.64 (P < 0.05, n = 62)35 predictive
			SF-36-PCS	r = 0.39 (P < 0.05, n = 61)35 predictive
			EQ-5D VAS	r = 0.22 (P ≥ 0.05, n = 64)35 predictive
Chronic (>6 mo)
3MWT	Body function	Motor function-lower extremity	Brunnstrom recovery stage	r = 0.49 (P < 0.05, n = 153)25
5MWT	Body function	Strength-knee extensorb	Dynamometer	r = 0.41 (P = 0.24, n = 10)28
6MWT	Body function	Aerobic capacity	Relative o 2	r = 0.66 (P < 0.05, n = 12)30
		Motor function-Lower extremity	Fugl-Meyer lower extremity score	ρ = 0.72 (P ≤ 0.001, n = 34, indoor 6MWT)40
				ρ = 0.74 (P ≤ 0.001, n = 34, outdoor 6MWT)40
				ρ = 0.80 (P < 0.01, n = 31)58
				ρ = 0.80 (P < 0.01, n = 31)69
			Chedoke-McMaster stroke assessment	r = 0.75 (P < 0.01, n = 25)63
		Strength-hip flexorb	Dynamometer	r = 0.40 (P = 0.02, n = 48)37
		Strength-hip extensorb		r = 0.40 (P = 0.002, n = 48)37
		Strength-knee flexorb		r = 0.20 (P = 0.2, n = 48)37
				r = 0.71 (P < 0.01, n = 50)33
		Strength-knee extensorb		r = 0.39 (P < 0.01, n = 63)38
				r = 0.40 (P = 0.01, n = 48)37
				r = 0.41 (P < 0.005, n = 63)32
				r = 0.57 (P < 0.001, n = 62)39
				r = 0.70 (P < 0.01, n = 50)33
		Strength-ankle dorsiflexorb	Dynamometer	r = 0.50 (P = 0.001, n = 48)37
			Peak torque	r = 0.79 (P ≤ 0.000, n = 62)59
		Strength-ankle plantarflexorb	Dynamometer	r = 0.43 (P < 0.05, n = 25)63
				r = 0.50 (P = 0.001, n = 48)37
			Peak torque	r = 0.35 (P = 0.005, n = 62)59
		Strength (arm, leg, hand, ankle/foot)b	Stroke impact scale-strength	r = 0.52 (P = 0.003, n = 30)43
	Activity	Balance	Berg balance scale	r = 0.67 (P ≤ 0.01, n = 21)46
				r = 0.69 (P < 0.001, n = 74)39
				r = 0.78 (P < 0.01, n = 25)63
				ρ = 0.67 (P ≤ 0.001, n = 34, indoor 6MWT)40
				ρ = 0.69 (P ≤ 0.001, n = 34, outdoor 6MWT)40
				ρ = 0.85 (P < 0.005, n = 63)32
		Mobility	Timed Up and Go	ρ = −0.96 (P < 0.01, n = 10)65
				r = −0.89 (P < 0.001, n = 50)31
		Physical activity	Physical activity scale for individuals with physical disabilities	ρ = 0.31 (P = 0.06, n = 40)51r = 0.32 (P < 0.05, n = 63)38
			ICF measure-activity	r = −0.32 (P < 0.05, n = 77)60
			Accelerometer activity kilocounts	ρ = 0.67 (P = 0.001, n = 40)51
			SAM-steps/d	r = 0.55 (P < 0.05, n = 17)57
			SAM-# steps at low ratec	r = 0.58 (P < 0.01, n = 49)47
			SAM-# steps at high ratec	r = 0.60 (P < 0.01, n = 49)47
			SAM-mean steps/d	r = 0.67 (P < 0.01, n = 49)47
			SAM peak activity index (steps/min)	r = 0.72 (P < 0.01, n = 49)47
			activPAL–mean minutes spent walking/standing/d	ρ = 0.48 (P = 0.051, n = 17)64
			IDEEA-time spent on ft/dd	r = 0.55 (P < 0.001, n = 42)45 predictive
			IDEEA-activity counts	r = 0.60 (P < 0.001, n = 42)45 predictive
			SAM-mean steps/d	r = 0.68 (P = 0.001, n = 19)49 predictive
		Stair function	Stair climbing-descent	r = −0.80 (P < 0.001, n = 50)31
			Stair climbing-ascent	r = −0.82 (P < 0.001, n = 50)31
		Strength-lower extremity	Five times sit-to-stand test	r = −0.60 (P < 0.001, n = 68)50
		Walking ability	Walk-12	ρ = 0.59 (P ≤ 0.01, n = 50)61
		Walking distance	12-min walk test	r = 0.97 (P < 0.01, n = 25)63
		Walking speed	8-m comfortable walk test	r = 0.92 (P < 0.01, n = 25)63
	Activity and participation	Activity and participation	ICF measure-total	r = −0.34 (P < 0.05, n = 77)60
	Participation	Mobility Participation	Stroke impact scale-mobility Stroke impact scale-participation	r = 0.72 (P < 0.001, n = 30)43r = 0.56 (P = 0.001, n = 30)43r = 0.53 (P < 0.01, n = 50)33 ρ = 0.43 (P = 0.082, n = 17)64
			ICF measure-participation	r = −0.31 (P < 0.05, n = 77)60
12MWT	Body function	Motor function-lower extremity	Chedoke-McMaster stroke assessment	r = 0.69 (P < 0.01, n = 25)63
		Strength-ankle plantarflexorb	KinCom	r = 0.35 (P ≥ 0.05, n = 25)63
	Activity	Balance	Berg balance scale	r = 0.80 (P < 0.01, n = 25)63
		Walking speed	8-m comfortable walk test	r = 0.91 (P < 0.01, n = 25)63

Abbreviations: ICF, International Classification of Functioning, Disability and Health; IDEEA, Intelligent Device for Energy Expenditure and Activity; NS, not significant; PCS, physical component summary; PF, physical function; SAM, StepWatch activity monitor; SSWS, self-selected walking speed; UE, upper extremity; VAS, visual analog scale; 3MWT, 3-minute walk test; 5MWT, 5-minute walk test; 6MWT, 6-minute walk test; 6MWD, 6-minute walk distance; 12MWT, 12-minute walk test.

^aNormalized values corrected for the influence of age, sex, height, and weight.

^bStrength of the paretic-side examined.

^cHigh rate defined as less than 60 steps per minute, low rate defined as less than 30 steps per minute.

^dSum of time spent walking, going up and down stairs, standing, and in sit-to-stand transition per day averaged over 2 days.

Table 4. - Summary of Reliability, Measurement Error, and Construct Validity Findings by Walk Test and Recovery Phase^a Poststroke

Walk Test	Reliability Coefficient (# Studies)			MDC90, m (Rounded) (# Studies)			Constructs Correlated With Walk Test Performanceb (# Correlations with P < 0.05)
	Acute	Subacute	Chronic	Acute	Subacute	Chronic	Acute	Subacute	Chronic
2MWT	1.00 (1)	1.00 (1)	0.98 (1)			11 (1)
3MWT			0.90 (1)						Motor functionc (1)
5MWT		0.97 (1)			24 (1)		Walk independence (1)
							Walk speed (1)
6MWT Walk with assist		0.97-0.98 (2) 0.80 (1)	0.96-0.99 (5)		39-52 (2) 34 (1)	28-42 (5)	Aerobic capacity (1)	Aerobic capacity (11)	Aerobic capacity (10)
							Mobility (2)	Mobility (2)	Mobility (2)
							Walk speed (2)	Walk speed (8)	Walk speed (4)
								Balance SE (1)	Balance SE (1)
									Motor functionc (5)
									Strengthc (15)
									Balance (6)
									Physical activity (11)
									Stairs (2)
									Walk distance (1)
									Walk ability-SR (1)
									Participation (5)
12MWT	0.68-0.71 (1)	0.68-0.71 (1)							Motor functionc (1) Balance (1)
									Walk speed (1)

Abbreviations: MDC₉₀, minimal detectable change at the 90% confidence level; SE, self-efficacy; SR, self-reported; 2-, 3-, 5-, 6-, 12MWT, 2-, 3-, 5-, 6-, 12-minute walk test.

^aResults from studies combining people in 2 phases of stroke recovery were listed under both phases.

^bConstructs with statistically significant correlations (P < 0.05) in 80% or more of correlations performed were listed.

^cThe lower extremities.

Sensitivity to Change

Sensitivity to change was reported in one study⁶² examining the 12MWT among participants in the acute and subacute phase poststroke. The standardized response mean for the 12MWT was 1.90 (ie, large^67,68).

DISCUSSION

This is one of the first comprehensive systematic reviews to synthesize research evidence describing the measurement properties and walk test protocols of time-limited walk tests in people with stroke. Findings support the excellent test-retest reliability of various 6MWT protocols in the subacute^41,42 and chronic^{30,31,42,56,65} phases of stroke recovery and the construct validity of 6MWT performance as a measure of functional walking capacity in people with acute,^29,34 subacute,^{29,31,32,34–36,39,41,46,48,52,54,55} and chronic^{30–33,35–40,43,45,46–52,54,55,57–61,63,69} stroke. Estimates of MDC₉₀ for the 6MWT range from 39⁴² to 52 m⁴¹ and from 28⁵⁶ to 42 m^30,31,42,65 in the subacute and chronic phases of stroke recovery, respectively. Few reports evaluating the 2-, 3-, 5-, and 12MWT poststroke were available. Preliminary evidence indicates that for the 2MWT, intrarater reliability is excellent in acute and subacute recovery phases,²³ test-retest reliability is excellent in the chronic phase,²⁴ and the estimated MDC₉₀ is 11 m in people with chronic stroke.²⁴ No studies examining the construct validity of the 2MWT poststroke were found. In people with chronic stroke, the 3MWT demonstrates excellent test-retest reliability with evidence of construct validity limited to one correlation with motor function.²⁵ The 5MWT has excellent test-retest reliability in people with subacute stroke,²⁷ an estimated MDC₉₀ of 24 m²⁷ (subacute stroke), and preliminary evidence of construct validity in the acute²⁶ recovery phase. Intra- and interrater reliability of the 12MWT is acceptable in people with acute stroke,⁶² and preliminary evidence supports construct validity in the chronic phase.⁶³ Across walk tests, only the 12MWT was evaluated for sensitivity to change and it was shown to be a responsive indicator in an acute and subacute stroke population.⁶²

In alignment with the KTA framework, synthesis and analysis of the extensive research evidence on the 6MWT in this review has yielded the following considerations that would inform the development of a clinical practice guide. First, a walk test protocol selected for use in a particular practice setting should have evidence of test-retest reliability obtained in patients with characteristics similar to those seen in the setting of interest. For example, although evidence of reliability of the 6MWT in the acute phase of stroke recovery is lacking, excellent reliability of the 6MWT has been observed in studies where evaluators provided physical assistance to walk as necessary.^41,56 These findings suggest that reliability of the 6MWT protocol used would be acceptable in the acute setting where 24% of patients who can walk require assistance.¹⁵

Second, it is essential to use a standard written protocol and documentation procedure if comparison of walk test performances over time or between patients within and across settings is desired. Although excellent reliability of the 6MWT has been reported for the subacute and chronic phases of stroke recovery, 6MWT protocols used in each phase varied in terms of walkway length and shape, location (indoor/outdoor), and encouragement, which can influence the distance walked and limit comparisons.^53,70–72 For example if a patient completed the 6MWT on a 10-m walkway in an acute care setting, and later walked 25 m further during the test on a 30-m walkway in a rehabilitation hospital, the improvement could have resulted from using a longer walkway⁵³ instead of an increase in walking capacity.

Third, review findings support inclusion of specific elements in a standardized protocol for 6MWT administration poststroke. An excellent basis for adapting a protocol to use poststroke is the recently updated 6MWT protocol for chronic respiratory disease⁷³ that is widely used in research and clinical practice.¹⁰ This protocol involves screening the patient for relative and absolute contraindications, and using a standardized set of equipment, instructions and encouragement statements delivered each minute, and a straight, 30-m walkway.⁷³ Patients are asked to wear comfortable clothing, supportive shoes, and their usual walking aids. Documentation of the distance walked, and use of any mobility devices, is recommended. Although 2 trials of the 6MWT is advised for people with respiratory disease, results from the current review show reliability is excellent conducting only 1 6MWT trial,^{30,31,41,42,56,65} and no practice trial^{31,41,42,56,65}; thus, a single 6MWT administration is recommended poststroke. We also advise recording the level of physical assistance provided, and the walkway length in communications of patient status to help colleagues who may readminister the test to better interpret the influence of these factors on observed change in performance. Once consensus on a standardized 6MWT protocol to use poststroke is reached, evaluating the reliability and measurement error of the protocol across the care continuum is advised to support its use.

The MDC₉₀ estimates obtained for the 6MWT in this review varied widely for subacute^41,42 (39-52 m) and chronic^{30,31,42,56,65} (28-42 m) phases of stroke recovery. This variability was likely due to differences in 6MWT protocols and stroke populations across studies that influenced reliability estimates (ICC values) and standard deviations used to calculate the MDC₉₀. A methodological issue that emerged from this review is that MDC₉₀ values derived using the standard deviation of performances and test-retest reliability estimates were consistently lower than values computed using within-subject error variance.²⁰ This finding highlights the importance of adequately reporting the statistical methods used to compute MDC values, and comparing values derived using the same method.

An MDC value is used to interpret whether the change in a patient's performance is sufficiently large to represent true change in ability.²⁰ The MDC at the 90% confidence level (MDC₉₀) means that 90% of truly unchanged patients will display random fluctuations in performance within the range of the MDC value.²⁰ Change must therefore exceed the MDC₉₀ value to be considered as “true change” in ability.²⁰ For example, the 6MWT MDC₉₀ value of 39 m for people with subacute and chronic stroke⁴² indicates that 90% of truly unchanged patients may display random fluctuations (improvements or deteriorations) in performance as large as 39 m. Therefore, a patient has to improve by greater than 39 m on the 6MWT to interpret the change as true improvement in walking capacity. Clinicians should consider whether sample size was adequate, defined as n ≥ 25 in this study, prior to selecting a 6MWT protocol based on reliability results, or an MDC value to interpret performance, and examine whether characteristics of study participants are similar to patients seen in clinical practice.

Review findings revealed gaps in the literature. Despite extensive evaluation of the 6MWT, no studies examining reliability and measurement error of the 6MWT in the acute phase and few studies in the subacute phase of stroke recovery were identified. The vast majority of the literature reporting associations between 6MWT performance and measures of physical capacity, physical activity, and participation involved people with chronic stroke. Evidence of reliability and validity of the 6MWT poststroke in acute and subacute care settings would help influence physical therapists to implement the test in these settings.¹ Finally, there is a general lack of research investigating the measurement properties of alternate time-limited walk tests. Consensus as to which time-limited walk tests should be used poststroke would guide the focus of future research.

Limitations of this review include the inability to include all constructs in validity studies or a more current review due to the extensiveness of the literature in this area. To optimize the comprehensiveness of our review of published validity evidence, we included any study examining associations between walk test performance and measures of targeted constructs.

CONCLUSIONS

Various 6MWT protocols demonstrate excellent test-retest reliability and yield estimates of measurement error in the subacute and chronic phases of stroke recovery. Evidence supports the construct validity of using the 6MWT, with a standardized administration protocol, in people with acute, subacute, and chronic stroke. Investigation of the measurement properties of the 2-, 3-, 5-, and 12MWT is limited. Methodological weaknesses in the literature highlighted in this review will inform future research. Considerations for advancing implementation of the 6MWT as a recommended measure of functional walking capacity poststroke are provided.