Measurement of Energy Cost by the Physiological Cost Index in Walking After Stroke

Methods 

Persons with stroke recruited from patient records in a rehabilitation department were asked to volunteer for the study. Our inclusion criteria were first time stroke according to World Health Organization criteria at least 6 months previously, 18 to 65 years of age, hemiparesis, a stable heart condition, and walking ability without manual assistance for 5 minutes (if necessary with a walking aid or orthosis). Exclusion criteria were severe cardiac disease or arrhythmia, pain during walking, walking impairment other than stroke-induced, inability to understand information or follow instructions, and severe discomfort while wearing a face mask. We included 20 subjects (17 men, 3 women). They were 30 to 63 years of age (median, 54y) with a mean body mass index (BMI) of 25kg/m2. Eleven subjects had a cerebral infarction and 9 had a cerebral hemorrhage. Twelve had lesions in the right hemisphere and 8 in the left. Time elapsed from stroke ranged from 7 to 96 months (median, 19mo). Motor function of the lower-limb section of the Fugl-Meyer Assessment35 scored a median of 22 (normative motor control, 34). Thirteen subjects used a cane and 15 an AFO. Ten of the stroke subjects were on antihypertensive medication; 5 had a β-blocker, 4 a calcium channel blocker, 5 an angiotensin-converting enzyme inhibitor, 1 an angiotensin II antagonist, 1 an α-blocker, and 4 had a combination of 2 or 3 of these medications. Eleven of the stroke subjects equipped with an AFO formed a subgroup additionally assessed without the AFO.

As a reference group, we recruited 16 healthy volunteers (11 men, 5 women) without walking impairments or known cardiac disease, 33 to 64 years of age (median, 49y), with a mean BMI of 24kg/m2 from the staff at the clinic.

Participants received verbal and written information and gave their signed consent. The study was approved by the Ethics Committee at Göteborg University.

Results 

Data from the 2 test sessions are presented in table 1. Energy cost was approximately double for the stroke subjects compared with that of the healthy subjects, reflected by PCI and V̇o2 levels. There were differences between mean treadmill walking speeds which, in addition to V̇o2 per time unit, were lower for the stroke group (table 1). The stroke group had higher V̇o2 per unit distance, PCI, and perceived exertion than the reference group, although walking heart rate or RER did not differ between groups for any of the sessions (see table 1). The significance test showed no differences between the 2 test sessions for any of the variables with the exception of resting heart rate in the healthy group, which was lower at retest (P=.027). No differences in heart rate, PCI, or V̇o2 were found for the stroke subjects with respect to lesion site, diagnosis, time elapsed from stroke onset, or administration of antihypertensive medication.

Table 1. Data From First and Second Test Sessions for the Stroke and Healthy Reference Groups

	Variable	Test	Stroke	Healthy	P⁎
	Resting heart rate (beats/min)	1	73±10	73±8	NS
		2	73±8	69±6	.040
	Walking heart rate (beats/min)	1	91±16	90±9	NS
		2	91±16	88±9	NS
	PCI (beats/m)	1	0.76±0.50	0.28±0.08	<.001
		2	0.76±0.59	0.32±0.10	<.001
	V̇o2 (mL·kg−1·min−1)	1	9.1±2.1	11.5±2.0	.004
		2	9.1±2.3	11.5±2.0	.002
	V̇o2 (mL·kg−1·m−1)	1	0.41±0.21	0.19±0.04	.001
		2	0.40±0.19	0.19±0.04	<.001
	Perceived exertion†	1	3	1	.001
		2	2	1	.004
	Walking speed (m/s)	1, 2	0.48±0.28	1.01±0.15	<.001

NOTE. Values are mean ± standard deviation. Abbreviation: NS, not significant.

⁎ Difference between stroke and healthy subjects (Mann-Whitney U test). † Median score on the Borg CR-10 (perceived exertion range, 0−10).

The ICC coefficients are shown in table 2. Visual analysis of Bland-Altman plots for the differences between first and second test sessions (Fig 1, Fig 2) showed a normal distribution. A larger error interval was seen for PCI than for V̇o2 (in mL·kg−1·m−1) (see table 2, Fig 1, Fig 2), particularly for the stroke group. The regression analysis for PCI showed that the model including age, sex, group assignment, and V̇o2 could explain 53% of the variation (table 3). Because site of lesion, stroke diagnosis or elapsed time from stroke onset were not significant in the bivariate analysis, they were not included in the regression model.

Table 2. Differences and Correlations Between Test Sessions 1 and 2 in the PCI and V̇o2

	Test	Subjects	Mean Difference	95%CI	95%LOA	ICC2,1
	PCI (beats/m)	Stroke(n=20)	−.009	−.142to.125	.55to−.57	.86
		Healthy(n=16)	−.037	−.083to.009	.13to−.21	.57
	V̇o2(mL·kg−1·m−1)	Stroke(n=20)	.012	−.008to.033	.10to−.08	.98
		Healthy(n=15)	.003	−.008to.014	.04to−.04	.87

Abbreviations: CI, confidence interval; LOA, limits of agreement.

View Large Image Download to PowerPoint

Fig 1. Individual differences in PCI (in beats/m) between the 2 test sessions in the (A) stroke and (B) healthy reference groups, respectively. Abbreviation: SD, standard deviation.

View Large Image Download to PowerPoint

Fig 2. Individual differences in V̇o2 per unit distance between the 2 test sessions in the (A) stroke and (B) healthy reference groups, respectively.

Table 3. Regression Analysis of the PCI

	Variable	Parameter Estimate	95% CI	P
	(constant)	−0.202	−0.900to0.496	.560
	Sex	0.072	−0.288to0.432	.686
	Age	0.001	−0.014to0.017	.853
	Group	0.060	−0.293to0.412	.731
	V̇o2(mL·kg−1·m−1)	1.862	0.934to2.791	.000

NOTE. R2=.534 (adjusted R2=.474).

The subgroup of stroke subjects measured both with and without an AFO had a mean PCI value of .75 (median, .62) without and .73 (median, .66) with the AFO, which was not significantly different (P=.722). Oxygen costs were .52 (median, .51) and .46 (median, .47) mL·kg−1·m−1, which were significantly lower with the AFO (P=.032). The perceived exertion did not differ between the 2 test conditions.

Discussion 

In the present study, the reliability and validity of the PCI in a clinical sample of stroke subjects were found to be limited when compared with direct measurement of V̇o2.

No systematic differences between 2 test sessions were detected in either PCI or V̇o2. The ICCs for PCI were high for the stroke group and moderate for the healthy subjects. The ICC for PCI in the stroke subjects was comparable with that in another study of adults with brain injuries.24 For V̇o2, the ICC was very high in the stroke group and high in the healthy subject group. A correlation coefficient is influenced by the range of variation between subjects, so the high correlation in the stroke group might have been false. Analysis of test-retest differences with Bland-Altman plots and the method of 95% limits of agreement showed small mean differences but wide error intervals in the case of PCI, which implies that, compared with V̇O2, the measurement precision was poorer for PCI. This was seen in both the stroke and healthy groups but was more obvious among those with stroke. The greater dispersion seen in the stroke group could be explained by their heterogeneity in neurologic deficit and walking speed, which was reflected in the stroke participants’ larger spread in the V̇o2 measures as well. Surprisingly, the increase in heart rate between rest and work did not differ between groups. The small increase could be explained by the low walking speed in the stroke group, because heart rate is related to walking speed. Heart rate increase could be suppressed by antihypertensive medication, but no differences were found between subjects with and without these drugs. Impaired heart rate variability (HRV)31, 32 reflecting autonomic cardiac dysfunction may theoretically be another explanation. Structural lesions in the nervous system resulting in changes in vagal and/or sympathetic activity may contribute to an altered heart rate response to exercise. Measures of HRV have been shown to be lower than in control subjects, mainly in the early phase after stroke, but even 6 months later.32, 34 In the present group, however, more than 6 months had passed. Signs of cardiovascular comorbidity in the acute or later stage have been reported in 75% of stroke patients.5 There are indications that right-sided cerebral lesions may involve reduced HRV,31 but HRV was not examined in the present study. In the small subgroups, no lesion side differences were found for heart rate response to exercise.

The aim of the present work was to investigate the PCI method in a clinical situation with as much control as possible without interfering with prescribed medications. A limitation in the study design is the lack of control of the role of antihypertensive medication. A purely physiologic study of the relationship between PCI and V̇o2 would require a population not taking medication that could affect the heart rate response. Additionally there were no data on aerobic fitness collected in this study group. Heart rate also may have been affected by other factors such as stress, environmental conditions, or other comorbidity.

Burridge et al9 reported a significant reduction in PCI values in stroke subjects that walked with functional electric stimulation. Olney et al29 showed a significant reduction in PCI in a group of stroke subjects after an exercise program. In the present study, we were not able to detect any reduction in the PCI when an AFO was applied, in contrast to the V̇o2 measure, which differed significantly. The inability to detect significant differences with PCI may be explained by the large variation seen in the test-retest analysis. Our sample for this comparison consisted of only 11 subjects and the size of this difference was not clinically relevant because no change in the perceived exertion was reported.

PCI values from both healthy and stroke subjects in the present study were within the range reported by other authors.9, 13, 18, 19, 21, 24, 29, 39 However, reliability studies carried out in healthy persons cannot be generalized to patient populations, where variability is larger. Several authors17, 18, 20, 22 have questioned the PCI as an outcome measure in group comparisons owing to the large variability. Caution must be taken in interpreting PCI measurements in patients, and it should not be the only outcome measure in a research study. PCI might be useful on the individual level as a clinical tool to obtain some information on energy cost.17

It may be that the increase in energy cost in persons with hemiparesis is not high enough to be measurable by PCI because of a large variability or low responsiveness to change. The energy cost of walking after stroke is about twice as high as normal. Perhaps PCI would be more suitable in patient groups with even higher levels of energy cost, for example, in persons with paraparesis or after lower-limb amputation. Presenting the level of change by percentage may be very misleading, depending on the initial level; a 50% change for a healthy person is not so remarkable, but it might be extraordinary if the initial level is increased many times. The clinically significant size of a change in PCI is not known.

Another limitation of the present study is that the measurement equipment required testing on a treadmill. The idea of PCI is that a self-selected speed is used. In our study, each person’s self-selected speed was used but then held constant at both sessions, which might have resulted in too high a PCI reproducibility. Our results therefore might reflect the reproducibility of heart rate more than PCI because no change in walking speed was allowed between the 2 test sessions. There was a high variability of gait speeds in the stroke group. The reliability of the measurement of self-selected walking speed was not determined in the present study but high retest reliability for walking speed and distance has previously been shown in healthy persons as well as stroke patients.24, 40, 41, 42 A handrail support was allowed for maintaining balance on the treadmill, but there was no control as to what extent this differed between test sessions. Measurement of walking at free speed on the ground would have been preferable because treadmill walking might involve higher energy expenditure, at least in the elderly,43 although conflicting opinions on this matter exist.44, 45, 46

Further studies on the reliability and responsiveness to change of the PCI after stroke are needed. There is also a need to investigate how large a change in the PCI must be to reflect a clinically important difference.

Conclusions 

The PCI has limited reliability and validity as a measure of energy cost after stroke due to large individual variation between test and retest in comparison with measures of V̇o2. It may, however, be useful as a simple measure for patients in a clinical situation. Caution must be exercised when drawing conclusions considering the large variability between test and retest. Measurement of V̇o2, on the other hand, is more precise and can be recommended for studies of energy cost.