Abstract
Background Restoration of walking capacity, as reflected by walking speed and walking distance, is a primary goal after stroke. Peak aerobic capacity (peak oxygen consumption [V̇o2peak]) is suggested to be correlated with walking capacity after stroke. Although the strength of this correlation is unclear, physical therapy programs often target walking capacity by means of aerobic training.
Purpose The purpose of this systematic review was to summarize the available evidence on the correlation between V̇o2peak and walking capacity.
Data Sources The databases MEDLINE, CINAHL, EMBASE, Cochrane Library, and SPORTDiscus were searched up to May 2014.
Study Selection Cross-sectional studies reporting correlation coefficients between V̇o2peak and walking capacity in stroke were included, along with longitudinal studies reporting these correlation coefficients at baseline.
Data Extraction The methodological quality of the studies was assessed using a checklist of 27 items for observational research. Information on study design, stroke severity and recovery, and assessments and outcome of V̇o2peak and walking capacity, as well as the reported correlation coefficients, were extracted.
Data Synthesis Thirteen studies involving 454 participants were included. Meta-analyses showed combined correlation coefficients (rɱ) for V̇o2peak and walking speed and for V̇o2peak and walking distance of .42 (95% credibility interval=.31, .54) and .52 (95% credibility interval=.42, .62), respectively.
Limitations The studies included in the present review had small sample sizes and low methodological quality. Clinical and methodological diversity challenged the comparability of the included studies, despite statistical homogeneity. Relevant data of 3 studies could not be retrieved.
Conclusions The strength of the correlation of V̇o2peak with walking speed was low and moderate for V̇o2peak and walking distance, respectively, indicating that other factors, besides V̇o2peak, determine walking capacity after stroke.
Improving walking capacity is often a primary goal in rehabilitation after stroke.1,2 In stroke, walking capacity reflects the autonomy in walking enabling daily life mobility,3 which can be expressed in walking distance and walking speed.4
Recent large-scale intervention studies have demonstrated postrehabilitation mean walking distances achieved in 6 minutes ranging from 168 m5 to 416 m6 in individuals after stroke. These mean walking distances are significantly less than the mean value for healthy populations, which varies between 510 and 638 m.7 A meta-analysis on the effects of rehabilitation on walking speed after stroke, including 28 trials,8 showed mean values of walking speed at baseline in individuals with stroke varying between 0.11 and 1.20 m/s, as opposed to 1.20 to 1.46 m/s in healthy elderly adults.9
It has been suggested that walking capacity in individuals after stroke is positively associated with motor functions such as lower limb strength,10–12 balance,13–15 and cognitive functions.16 Similarly, peak aerobic capacity (V̇o2peak) has been suggested to be an important indicator of walking capacity after stroke.15,17–19 To date, statistically significant positive correlation coefficients (r) have been found between V̇o2peak and walking distance15,17 and walking speed18,19 after stroke. However, the magnitude of reported correlation coefficients between V̇o2peak and walking capacity varies considerably. Some studies reported low correlation coefficients of .2918 and .37,20 whereas other studies showed high values of .7121 and .74.17 Despite this broad range of reported correlation coefficients, many physical therapy programs in stroke target walking capacity by means of aerobic training. A number of recent reviews22–24 suggest that these programs have positive effects, which appears to confirm the relationship between V̇o2peak and walking capacity. However, these reviews22–24 were not aimed at establishing an overall conclusion about the correlation between V̇o2peak and walking capacity. Moreover, only a few of the included studies reporting on the effects of aerobic training on walking capacity actually measured both V̇o2peak and walking capacity.24 As a consequence, the true strength of the correlation between V̇o2peak and walking capacity remains unclear.
A clearer perspective on the strength of the correlation between V̇o2peak and walking capacity after stroke may add to the rationale behind the incorporation of aerobic exercise into rehabilitation programs. Therefore, the aim of the present systematic review was to summarize the available evidence on the magnitude of the reported correlation coefficients between V̇o2peak and walking capacity (ie, walking distance and walking speed) in individuals after stroke.
Method
The present study was a systematic review of the available literature. The PRISMA statement was followed for reporting items of this systematic review.25
Data Sources and Searches
In the initial computerized search, the databases MEDLINE, CINAHL, EMBASE, Cochrane Library, and SPORTDiscus were searched for relevant publications. Included publications were screened by hand to identify any additional publications for inclusion. The search was completed on May 30, 2014.
The search terms “stroke,” “aerobic capacity,” and “walking capacity” were used to develop a PubMed search string to search MEDLINE, which was afterward adapted to the search machines of the other databases. All known synonyms and related terms of the search terms were collected, and, where available, MeSH headings, CINAHL headings, and Emtree headings were used in the search strategy. An information specialist (J.M.) was consulted to compile the search strings. Appendix 1 provides an expansion of all search terms entered in PubMed searching MEDLINE. The search strings used for the other databases are available on request from the corresponding author.
One researcher (J.O.) performed the data search in cooperation with the information specialist (J.M.). All retrieved citations were imported in RefWorks 2.0 (RefWorks, Bethesda, Maryland), after which the duplicates were removed.
Study Selection
The following selection criteria were used to identify the relevant publications. First, the study design concerned cross-sectional or longitudinal studies, randomized controlled trials, and reliability studies reporting relevant correlation coefficients between V̇o2peak and walking capacity (ie, walking distance or walking speed). Second, the population included in the studies concerned patients with stroke, as defined by the World Health Organization (WHO),26 over 18 years of age. Third, the investigated variables were walking capacity and V̇o2peak. Walking capacity was defined as “the degree of autonomy in walking, with or without the aid of appropriate assistive devices (such as canes or walkers), safely and sufficiently to carry out mobility-related activities of daily living,”3 expressed in walking distance or walking speed as measured in standardized circumstances.4 Peak aerobic capacity was defined as the highest oxygen uptake an individual attains during physical work using large muscles in the lower extremities (ie, during walking or cycling) while breathing air at sea level measured during standardized circumstances.27
The studies had to report a physically performed assessment of walking capacity, such as the Six-Minute Walk Test (6MWT) (m) or the Ten-Meter Timed Walk Test (10MTWT) (m/s), and V̇o2peak, using treadmill or bicycle ergometer protocols and breath-by-breath gas analysis equipment. Last, the publications were written in English, Dutch, German, or French.
All retrieved publications were screened with respect to the inclusion criteria, first on title and abstract and subsequently on full text. Two researchers (J.O. and I.vdP.) independently performed the screening and selection of articles. Throughout the selection process, all disagreement regarding inclusion was discussed until consensus with respect to the inclusion criteria was reached. If disagreement persisted, a third author (H.W.) was consulted.
Data Extraction and Quality Assessment
The following information was extracted: study ID (author and year); study design; source population and recruitment; number of participants; age; time since onset of stroke; stroke type, localization, and severity; assessment protocols; and mean values of V̇o2peak, walking speed, and walking distance and the correlation coefficients between walking capacity and V̇o2peak. Corresponding authors of included studies were contacted in case of inconclusive or incomplete data.
Two researchers (J.O. and I.vdP.) independently performed the quality assessment. The quality of the studies was assessed with a checklist to evaluate prognostic studies,28 which is based on the Strengthening Observational Studies in Epidemiology (STROBE) guidelines.29 The 27-item checklist addresses 6 major risks of bias: study participation, study attrition, predictor measurement, outcome measurement, statistical analysis, and clinical performance and validity. The checklist is available in Appendix 2. In case the included studies involved cross-sectional studies, the items “inception cohort” (D5), “information about treatment” (D6), “number of losses to follow-up” (A1), “reasons for loss to follow-up” (A2), “comparison of completers and noncompleters” (A4), and “appropriate end-points of observations” (O4) were considered as “not applicable.”
Each item was graded “positive,” “negative,” ”unknown/partial,” or “not applicable.” A positive grade was given in case of sufficient information, indicating low risk of bias, assigning 1 point to the item. A negative grade was given in case there was no information, indicating a high risk of bias, assigning 0 points to the item. An unknown/partial grade was given in case of insufficient information, leaving the risk of bias unknown, assigning a “?” to the item. A not applicable (NA) grade was given when the item was not applicable for the evaluated study.
Summing all items that were graded positive and dividing the sum by the total number of applicable items resulted in the total score. A study was considered to have a low risk of bias when it scored ≥75% of its maximum score; otherwise, the studies were considered to have a high risk of bias.28
With respect to the interrater reliability of the quality assessment of the studies, the percentage of agreement on the items and Cohen kappa (κ) were calculated. The Cohen kappa value was considered poor (≤0), slight (.0–.20), fair (.21–.40), moderate (.41–.60), substantial (.61–.80), or almost perfect (.81–1.0).30 Calculations were performed with IBM SPSS version 20 (IBM Corp, Armonk, New York).
Data Synthesis and Analysis
To explore possible publication bias, a funnel plot was made by plotting the correlation coefficients (r) against the number of participants in the study. Next, the symmetry of the plot was assessed visually, where the studies should be symmetrically distributed on both sides of the combined correlation coefficient line to indicate the absence of publication bias.31 The visual assessment was performed separately for the studies correlating V̇o2peak and walking speed and V̇o2peak and walking distance, respectively. A heterogeneity analysis was performed to determine statistical heterogeneity as a cause for asymmetry in the funnel plots,31 using the Higgins I2 test.32 As proposed by Higgins and colleagues,32 a value higher than 50% was considered an indicator of substantial heterogeneity. The I2 value was calculated from the Q statistic as proposed by Hunter and Schmidt33,34 from the sample sizes and effect sizes.
Two meta-analyses were performed, pooling the studies with respect to walking speed and walking distance. The combined correlation coefficient (rɱ) between walking capacity and V̇o2peak was obtained by using all of the reported correlation coefficients (ie, Pearson correlation coefficient [rp] and Spearman rank correlation [rs]), using the random effects method as described by Hunter and Schmidt.33,34 In case of a longitudinal design, only reported baseline data were used for the present review. The combined correlation coefficient (rɱ) was considered low (.26–.49), moderate (.50–.69), high (.70–.89), or very high (.90–1.00).35 Generalizability of the calculated value was estimated by a credibility interval using the variance in population correlations.34 The statistical significance of the difference between the combined correlation coefficient of V̇o2peak with walking speed and walking distance, respectively, was calculated using the Student t test (P≤.05). The statistical analyses were performed using Microsoft Office Excel 2013 (Microsoft Corp, Redmond, Washington), incorporating the calculations as proposed by Hunter and Schmidt.33,34
Role of the Funding Source
This study was funded by SIA RAAK International (project number: 2010-2-024 INT).
Results
Study Selection
The searches of the databases delivered a total of 1,613 citations, as shown in the flowchart (Fig. 1). After subtraction of duplicate records, 1,125 studies remained to be screened on title and thereafter 152 for screening on abstract. Four studies were excluded for language reasons, 25 were excluded on design, and 73 were excluded on outcome measures, and only an abstract was available for 5 studies. Forty-five citations remained for full-text examination. One study was excluded for nonavailability of relevant data.36 Fifteen studies were excluded for not reporting correlations between V̇o2peak and walking capacity, and 16 studies were excluded for not using physical assessments of performance. Finally, 13 studies were included in this review, all reporting correlations between V̇o2peak and walking capacity, as shown in Table 1.
Flowchart of publication selection.
Forest plot depicting effect sizes (r) for the association of peak oxygen consumption (V̇o2peak) with mean values of walking speed on the Ten-Meter Timed Walk Test (10MTWT) and walking distance measured with the Six-Minute Walk Test (6MWT) for individual studies. Error bars depict the 95% credibility interval.
Study Characteristicsa
Visual assessment of the funnel plots showed slight asymmetry (Figs. 3 and 4). Heterogeneity analysis showed an overall homogeneous sample for the studies concerning walking speed (I2=0%, Q=3.80, df=4, P=.63) as well as walking distance (I2=0%, Q=8.24, df=9, P=.51).
Funnel plot for the included studies concerning walking distance.
Funnel plot for the included studies concerning walking speed.
Study and Participant Characteristics
Table 1 shows that 10 of the 13 included studies were cross-sectional cohort studies.14,15,17,18,21,37–41 Three studies used other designs: 2 studies used a longitudinal design,19,20 and 1 study concerned the baseline analysis of a clinical trial.42 Of these 3 studies, only the correlation coefficients that were calculated from baseline assessments were used in the analyses of the present review.
A total of 454 participants (184 female, 270 male), with a mean age ranging from 5321 to 6839 years, were included in the present review. In 2 of the studies, the participants were less than 3 months poststroke19,41 (Tab. 1). Two studies did not report stroke type.15,38 The majority of the patients in the remaining studies (ie, 258 out of 335) had an ischemic stroke. Six studies reported stroke localization,14,19,20,39,40,42 showing 110 out of 236 participants sustained a left hemispheric stroke, 117 a right hemispheric stroke, and 9 had a bilateral hemispheric stroke (Tab. 1). Stroke severity was reported in 4 studies15,18,19,41 using the National Institutes of Health Stroke Scale (NIHSS), with scores ranging from mean of 2.8 to 4.9 points out of a maximum of 42 points. Two studies14,20 classified their sample according to the American Heart Association Stroke Outcome Classification. One study14 classified 70% and the other study20 classified 100% of their samples in categories II and III. Two studies20,41 used the Chedoke-McMaster Stroke Assessment of the lower extremities to determine stroke recovery. Eng et al20 reported a mean score of 9.4 (maximum score=14 points) 3.5 years poststroke. Tang et al41 reported a mean score of 5.1 (maximum score=7 points) 3 months poststroke. The Brunnström-Fugl-Meyer assessment (BFM) was used in 4 studies21,37,40,42 to assess motor recovery in the lower extremities. The scores ranged from a mean of 19.1 points at 12.1 months poststroke to a mean of 30 points at 62 months poststroke. Two studies17,39 reported mean total BFM scores of 7917 and 6839 points at 47.617 and 1839 months poststroke, respectively.
Outcome Assessments
Table 1 shows that V̇o2peak was assessed on a treadmill in 4 studies15,18,21,38 and that a bicycle ergometer was used in the remaining 9 studies. Mean V̇o2peak ranged from 10.738 to 2214 mL O2·kg−1·min−1. One study39 used a one-legged bicycle protocol. Eight studies14,15,17,19,20,38,40,41 used the guidelines of the American College of Sports Medicine (ACSM),43 and 2 studies37,42 followed the approach described by Åstrand and Rodahl44 to determine V̇o2peak. Two studies19,38 reported the determination of V̇o2peak during the last minute of the test. Two studies14,20 reported maximum oxygen consumption (V̇o2max) in all participants, and one study17 reported this measure in 3 out of 9 participants. Four studies14,19,21,38 reported respiratory exchange ratios (RERs), with mean values of 0.96,21,38 1.01,19 and 1.12.14 Three studies14,20,21 reported peak heart rate (HRpeak) as a percentage of predicted maximal heart rate (%HRmax), showing means of 77.8%,21 98.1%,14 and 94.7%.20
Walking speed was assessed in 7 studies.15,18,19,21,37–39 Two studies used a 30-foot (1 ft=0.3048 m) timed walk,15,38 one study used a 20-m timed walk,37 and the other 4 studies used the 10MTWT. Two studies used maximal gait speed,19,21 as opposed to self-selected gait speed in the other 5 studies. Mean walking speed varied from 0.4218 to 1.5219 m/s (Tab. 1). Walking distance was assessed with the 6MWT in 10 studies.14,15,17,20,21,37,39–42 Seven studies15,17,21,37,39–41 used a straight course of 30 m or 100 ft. One study42 used a straight course of 70 m, and 2 studies14,20 used rectangular courses of 42 m. All studies14,15,17,20,21,37,39–42 reported the maximal walking distance in 6 minutes, which ranged from a mean of 216.015 to 400.921 m. One study14 reported the walking distance adjusted to leg length.
Study Quality
Table 2 shows that all studies scored less than 53% (range=38%–53%) of the maximal methodological quality score and were classified as “high risk of bias.” The items that were most often scored negative were “source population and recruitment” (item D1), “important key characteristics” (item D3), “measurement of V̇o2peak valid and reliable” (item P2), all items on statistical analyses, and all items on clinical performance.
Quality Assessment of the Included Studies
Percentage of agreement on the individual items between the 2 raters was 91%, with a Cohen kappa of .60. Percentage of agreement on the qualification of risk of bias was 100%. Both raters scored all studies as having a high risk of bias.
Synthesis of Results
Correlations between V̇o2peak and walking speed.
Seven studies calculated the correlation coefficient between walking speed assessed with a short timed walk and V̇o2peak.15,18,19,21,37–39 Two studies showed a statistically nonsignificant correlation coefficient but did not report the values.21,37 Figure 2 shows the 5 studies15,18,19,39,40 reporting statistically significant correlation coefficients (r) ranging from .2918 to .54.15
Correlations between V̇o2peak and walking distance.
Figure 2 shows the 10 studies that calculated the correlation coefficient between walking distance assessed with the 6MWT and V̇o2peak.14,15,17,20,21,37,39–42 All studies demonstrated statistically significant correlation coefficients (r) varying between .4214 and .741,7 except for Eng et al20 and Carvalho et al,40 who reported nonsignificant correlation coefficients of .37 and .34, respectively.
Three studies14,15,41 conducted a multivariate analysis. The study by Patterson et al15 showed that V̇o2peak explained most of the variance (48%) in walking distance on the 6MWT. The study also explored the difference in explained variance in 2 subgroups of slow (<0.48 m/s) and faster (>0.49 m/s) walkers and showed that balance explained 42% of the variance in the slower walkers, whereas V̇o2peak explained 26% of the variance in the faster walkers. The other 2 studies14,41 did not show V̇o2peak to be a significant determinant of walking distance on the 6MWT.
Meta-analyses.
Figure 2 shows the meta-analysis of the correlation coefficients between V̇o2peak and walking speed. A combined correlation coefficient (rɱ) of .42 (95% credibility interval [95% CI]=.31, .54) was calculated. Figure 2 also shows the meta-analysis of the correlation coefficients between V̇o2peak and walking distance. A combined correlation coefficient of .52 (95% CI=.42, .62) was calculated. The difference between combined correlation coefficients of V̇o2peak with walking speed and of V̇o2peak with walking distance was not statistically significant (P=.61).
Discussion
This systematic review provides an overview of the currently available evidence for the strength of the correlations between V̇o2peak and walking capacity, expressed as walking speed or walking distance, after stroke. The results of the present study show a low positive combined correlation coefficient between V̇o2peak and walking speed and a moderate combined positive correlation between V̇o2peak and walking distance that were both statistically significant.
These findings suggest that other factors, such as age, balance, stroke severity, or lower extremity muscle strength, may influence the correlation between V̇o2peak and walking capacity. In addition to the positive correlations between V̇o2peak and walking capacity, 4 of the included studies14,15,18,40 reported significant positive correlation coefficients between balance and walking capacity (r=.38–.85). Two studies37,40 reported significant positive correlation coefficients between stroke severity and walking capacity (r=.59–.72), and 3 studies14,15,39 reported significant positive correlations between knee extensor muscle strength and walking capacity (r=.18–.60). These reported correlation coefficients display a similar broad range and similar values as those between V̇o2peak and walking capacity. Additionally, the highest correlation coefficients (r≥.60) were reported in the studies using younger populations (mean age≤56 years),17,21,37,42 suggesting that age, as might be expected, may have influenced the correlation between V̇o2peak and walking capacity. All of these factors—age, balance, stroke severity, and lower extremity muscle strength—may have had an impact on the reported correlation coefficients and partly account for the broad range of correlation coefficients found in the present review.
Unfortunately, most studies included in this systematic review were limited to bivariate analyses, enabling only restricted insight into factors influencing the correlation between V̇o2peak and walking capacity. Only 3 of the included studies14,15,41 applied multivariate analyses to identify the determinants of walking capacity, unfortunately displaying disparate results. The first of these studies14 reported balance as the most important determinant for walking distance, explaining 66.5% of the variance of outcome. The second study15 showed V̇o2peak to be a significant predictor for performance on the 6MWT. This study15 furthermore reported that the variance of the 6MWT scores explained by V̇o2peak might differ in subpopulations, as balance was the strongest predictor in patients with slower walking speeds (<0.48 m/s), whereas V̇o2peak was the strongest predictor in those with faster walking speeds (>0.49 m/s). The last of these 3 studies41 identified fast walking speed as the main determinant for the 6MWT, explaining 65.4% of the variance of outcome.
The high predictive validity of walking speed for outcome of walking distance41 also may explain the absence of statistical significance with respect to the combined correlation coefficient of V̇o2peak with walking speed compared with that of V̇o2peak with walking distance. This finding suggests that, in individuals with stroke, the contribution of V̇o2peak is similar in walking speed, mostly assessed with short walks, and walking distance, mostly assessed with longer walks. This similarity indicates that both outcomes have a common underlying construct in patients with stroke, which is in line with the findings of earlier studies.45,46 Physiologically, the expectation would be that there is a significantly stronger correlation of V̇o2peak with walking distance than with walking speed, as a short walk would engage anaerobic metabolism, whereas a longer walk would engage aerobic metabolism.47 However, this expectation was not confirmed in the present study.
The clinical and methodological variability of the included studies appear to be substantial. Time since stroke was diverse. Two studies19,41 had samples of survivors of stroke less than 3 months poststroke, and 2 studies14,40 had samples more than 5 years poststroke, reflecting clinical diversity. Overall, however, mild to moderate stroke severity and modest to good recovery were reported, suggesting a similar level of functioning between the samples.
Concerning the methodological variability, the assessment of walking capacity also displayed some diversity, specifically in the courses used in the 6MWT. According to American Thoracic Society guidelines,48 the effects of the length of the course may not affect the outcome as long as it measures between 50 and 164 ft and presents a straight line. Three studies14,20,42 deviated from this recommendation. Furthermore, one study14 reported the results of the 6MWT adjusted for leg length, which may have affected the correlation. However, a post hoc sensitivity analysis without these studies14,20,42 showed a nonsignificant increase of the combined correlation coefficient (rɱ) of .57 (95% CI=.45, .67). Likewise, the assessment of V̇o2peak displayed diversity, as 4 studies used a treadmill protocol,15,18,21,38 whereas all other studies used a bicycle protocol.
It might be expected that the studies using a treadmill protocol to assess V̇o2peak would report a stronger correlation with walking capacity, as both assessments concerned walking protocols. However, this stronger correlation was not found in the present review. Furthermore, all except for 2 studies14,20 reported V̇o2peak, which reflects the highest amount of oxygen consumption attained during an exercise test but does not necessarily define the highest value attainable by the individual.49 Although the majority of included studies reported the use of guidelines,43,44 there was little information on the exact criteria used to determine V̇o2peak. Moreover, only 5 studies14,19,20,21,38 reported RER or %HRmax, which gives insight into the participants' effort during the assessments. The lack of information on both the exact criteria to determine V̇o2peak and the participants' effort during the assessments, as well as the use of different protocols, presents a challenge to the comparability of the studies' reported values of V̇o2peak. The differences in protocol and possibly participants' effort, in part, may explain the broad range of reported correlation coefficients.
In addition, the findings of the present study may be related to the lack of large-scale cohort studies affecting the precision of claimed correlation coefficients between walking speed and distance with V̇o2peak. Half of the included studies had small sample sizes (30 or fewer participants), which challenges statistical power and representativeness of the sample and could lead to an overestimation of the combined correlation coefficient, as the highest correlation coefficients were found in the smallest studies.17,20,21,42
Despite the displayed clinical and methodological diversity, the included studies presented a homogeneous sample, according to statistical testing, indicating that the studies were comparable. This finding suggests that the combined correlation coefficients are a true representation of the correlations between V̇o2peak and walking capacity in patients after stroke.
The methodological quality of all included studies was assessed as low, which although allowing for the pooling of the results of the included studies, challenges the strength of the evidence. Leaving out the studies that scored lowest in methodological quality of the meta-analyses only minimally altered the combined correlation coefficients between V̇o2peak and walking speed (rɱ=.41; 95% CI=.31, .50) and between V̇o2peak and walking distance (rɱ=.54; 95% CI=.46, .61). This finding did not change the interpretation of the strength of the found correlations, and the absence of a statistically significant difference between both combined correlation coefficients remained.
Limitations of the Study
First, the assessment of methodological quality was based on recent recommendations for prognostic research as well as criteria used in previous scoring lists for assessment of prognostic stroke research, as specific quality assessments for cross-sectional studies are lacking. The assessment of methodological quality was performed strictly, which may have underestimated the quality of the studies. For example, “source population and recruitment” (item D1) and “important key characteristics” (item D3) were graded positive in cases where the information matched exactly all criteria for the item. The overall negative grading of “measurement of V̇o2peak valid and reliable” (item P2) was related to the lack of information on the effort of the participants during the assessments. However, the quality assessment provided a good insight into the strategies used to prevent bias and confounding.
Second, relevant data of 3 studies could not be retrieved, indicating data availability bias. Two studies,21,37 which were included for reporting a correlation coefficient between V̇o2peak and walking distance, concluded that there was a nonsignificant correlation between V̇o2peak and walking speed. Unfortunately, they did not report the correlation coefficient. A third study36 reported correlation coefficients between both walking speed and distance and age-adjusted V̇o2peak. However, these studies had small sample sizes varying from 821 to 2137 participants, suggesting that these correlation coefficients may have had only a minor impact on the found combined correlation coefficient.
Third, despite a sensitive search, publication bias may still be present because of poor indexation of the literature reporting observational studies and because only published studies were considered. Visual inspection of the funnel plots showed slight asymmetry, suggesting the presence of publication bias. Asymmetry also could be explained by heterogeneity in study methods.31 However, as statistical testing showed homogeneity of the included sample, the asymmetry of the funnel plots is likely to be explained by the presence of publication bias.
Finally, none of the described methods to calculate combined correlation coefficients are completely suitable for a small number of studies, and the Hunter-Schmidt method tends to underestimate the combined correlation coefficient33,34 a little (ie, less than .011).34 However, Field and Gillett34 point out that in a Monte Carlo simulation, the bias was negligible and produced accurate estimates of the population effect size. This finding indicates that the calculated combined correlation coefficient in the present study is probably accurate.
Future Directions
Future observational research should follow the STROBE statements to increase methodological quality and aim at conducting larger studies, enabling multivariate analyses to reveal to what extent V̇o2peak can explain walking capacity. Balance and stroke severity should be taken into the equation as well as age. Physiological reserve, defined as oxygen consumption during walking relative to V̇o2peak,50 also may be considered, as it was shown that reduced balance51 and motor control52,53 increase oxygen requirements of walking in stroke. Although oxygen consumption during walking in stroke is minimized by means of reduction of walking speed,54 it can still remain at a high percentage of V̇o2peak, as the latter is reduced in stroke50 and decreases with age.47 Therefore, the physiological reserve, as it depends on oxygen consumption during walking as well as V̇o2peak, may be more strongly related to walking capacity than V̇o2peak itself. Consequently, in future research, exercise physiological variables such as oxygen uptake, RER, or heart rate during the walking tests and during maximal exercise testing should be assessed and reported. Reporting those variables gives insight into the effort during maximal exercise testing, allowing calculation of the physiological reserve and increased comparability between studies.
Implications of the Study
Although the results of the present study are to be considered carefully, a positive low to moderate correlation between walking capacity and V̇o2peak is suggested. In physical therapy interventions aimed at improving walking capacity after stroke, therefore, it appears legitimate to address aerobic capacity. However, as other factors (eg, age, balance, stroke severity, lower extremity muscle strength) are likely to affect the relationship between aerobic capacity and walking capacity, a multifactorial approach appears to be the most efficient.
Appendix 1.
Search String in PubMed
Appendix 2.
Quality Assessmenta
a Y (positive)=1 point, N (negative)=0 points. ?=unknown or partial, NA=not applicable, SD=standard deviation, CI=confidence interval, IQR=interquartile range, SE=standard error, OR=odds ratio, RR=relative risk, HR=hazard ratio, ROC=receiver operating characteristic, PPV=positive predictive value, NPV=negative predictive value.
Footnotes
Ms Outermans, Dr van de Port, Dr Wittink, and Dr Kwakkel provided concept/idea/research design. All authors provided writing. Ms Outermans and Dr van de Port provided data collection. Ms Outermans, Dr van de Port, and Dr de Groot provided data analysis. Dr van de Port provided project management. Dr Wittink provided fund procurement and facilities/equipment. Dr van de Port and Dr de Groot provided consultation (including review of manuscript before submission). The authors thank Jurgen Mollema, information specialist at the Hogeschool Utrecht University of Applied Science, and Cas Kruitwagen, statistician at the Hogeschool Utrecht University of Applied Science, for their assistance.
This study was funded by SIA RAAK International (project number: 2010-2-024 INT).
- Received March 3, 2014.
- Accepted December 21, 2014.
- © 2015 American Physical Therapy Association
References
Issue highlights
