Physiotherapy 97 (2011) 244–249

A plantar flexion device exercise programme for patients with peripheral

arterial disease: a randomised prospective feasibility study
Niki Tebbutt a , Lesley Robinson a , Jane Todhunter a , Leon Jonker b,∗
a North Cumbria University Hospitals NHS Trust, Cumberland Infirmary, Carlisle, UK
b Faculty of Health and Wellbeing, University of Cumbria, Carlisle CA1 2HH, UK

Abstract
Objectives To determine if the use of a plantar flexion device (Step It pedal) in a newly developed exercise programme is of benefit to patients
with peripheral arterial disease.
Design Prospective feasibility trial with patients randomised to either standard care or the Step It exercise programme plus standard care.
Setting Physiotherapy Department at Cumberland Infirmary, Carlisle, UK.
Participants Patients were identified from the vascular team’s referral list. In total, 42 patients agreed to take part; 18 in the control group
and 24 in the intervention group.
Interventions Eligible participants were randomised and received either standard care or took part in a plantar flexion resistance exercise
programme, involving the Step It pedal, for a period of 12 weeks.
Main outcome measures Maximum walking distance, claudication distance and ankle brachial pressure index.
Results Eighty-three percent of patients completed the study. Improvements in median distance to claudication symptoms and maximum
walking distance were observed in the intervention group but not in the control group. Nine out of 15 (60%) participants in the control
group and 14 out of 20 (70%) participants in the intervention group improved their walking distance. Ankle brachial pressure index remained
virtually unchanged in both groups.
Conclusions Due to the variability of patients’ fitness in the sample, it cannot be concluded whether use of the Step It pedal has additional
benefits to patients over standard care. However, the study completion rate implies that patients with peripheral arterial disease are receptive
to undertaking exercise programmes.
© 2010 Chartered Society of Physiotherapy. Published by Elsevier Ltd. All rights reserved.

Keywords: Peripheral arterial disease; Plantar flexion; Exercise; Claudication

Introduction arterial disease can be categorised using the Fontaine classi-

fication according to the absence or presence of intermittent
Peripheral arterial disease is a common progressive disor- claudication symptoms. In the USA alone, approximately
der of the vasculature. The underlying aetiology of peripheral 6% of the population are affected by this disease [4]. A UK
arterial disease is atherosclerosis, and therefore patients are study involving adults aged 55 to 74 years found that 4.5%
at high risk of associated cardiovascular diseases such as experienced intermittent claudication symptoms [5].
myocardial infarction and stroke [1,2]. Narrowing of the Patients who receive support, training and education about
arteries leads to reduced oxygen supply, thereby resulting exercise improve their walking distance by, on average, 150%
in intermittent claudication symptoms (i.e. exercise-induced [6,7]. Most previous research studies applied programmes
pain in the calves, thighs or buttocks), limited capacity to involving weight-bearing or aerobic exercise (e.g. walking,
exercise and increased risk of tissue loss [3]. Peripheral rowing and cycling) to try and improve the maximum walk-
ing distance that patients can reach. Due to the increased risk
∗ of cardiovascular accidents in patients with peripheral arte-
Corresponding author. Tel.: +44 01228 814689; fax: +44 01228 814689.
E-mail addresses: leon.jonker@cumbria.ac.uk, rial disease, the underlying atherosclerosis and the aversion
leon.cumbria@yahoo.co.uk (L. Jonker). to strenuous exercise by most patients, an alternative exer-

0031-9406/$ – see front matter © 2010 Chartered Society of Physiotherapy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.physio.2010.08.009
 N. Tebbutt et al. / Physiotherapy 97 (2011) 244–249 245

cise programme that does not involve high-impact aerobic was obtained – and visible to researchers beforehand – for
exercise regimes may be more appropriate than the afore- the whole sample using a freeware randomisation program
mentioned cycling and rowing programmes. (http://www.randomizer.org). The corresponding author (LJ)
Supervised exercise programmes have been shown to ben- generated the randomisation sequence, LR and JT enrolled
efit patients [8,9]. One recent study by McDermott et al. patients, and NT measured the study outcomes.
showed that supervised lower extremity resistance training
has significant benefits over normal management involving Interventions
education and advice on diet and exercise [10]. A 6-month
programme resulted in improved maximum treadmill walk- Participants were randomised to either the control group
ing time, although 6-minute walking distance did not improve (standard care) or the intervention group (standard care plus
over that of the control group. The lower extremity exercises use of the Step It pedal) for a period of 12 weeks. Neither par-
included repetitions of knee extensions, leg presses and leg ticipants nor researchers were blinded regarding the allocated
curls. intervention. All participants initially attended one exercise
This pilot clinical research study applied the Step It rocker class in the hospital for baseline measurements, and then con-
pedal (Step It System AB, Saltsjöbaden, Sweden) for the first tinued their exercises (i.e. plantar flexion using the Step It
time in patients with peripheral arterial disease. This is a small pedal) and care programme unsupervised at home thereafter.
device, very similar to the pedal used to operate a bass drum Patients in both groups were advised to walk to their max-
on a drum kit, which is easy to use from a seated position imum walking distance each day and to attempt to increase
and was first devised to help alleviate the risk of ‘economy this distance as they were able. The second study appointment
syndrome’ (i.e. deep vein thrombosis) for travellers on long in the hospital took place 12 weeks after patient enrolment. In
haul flights. Patients may be receptive to this form of exercise addition to the standard care received by the control group, the
in addition to walking, especially if they are elderly, frail, have subjects in the intervention group were also asked to under-
a fear of falling or have an aversion to the idea of ‘exercise’. take exercises at home using the Step It rocking pedal. These
The aim of this initial randomised, controlled, prospective exercises consist of lower limb exercise training (resisted
feasibility study was to determine the efficacy of the Step plantar flexion) whilst seated. The resistance of the pedal is
It pedal by measuring patients’ maximum walking distance equivalent to approximately 6 kg. The exercise sessions were
as well as ankle brachial pressure index. Both the control performed three times per week for 12 weeks with the fol-
group and the intervention group received advice on diet and lowing pattern: 2 minutes exercise/2 minutes rest, 10 times,
exercise, and were monitored for physical parameters. to equal 20 minutes of exercise in total. The patients were
shown how to use the Step It pedal at the baseline appoint-
ment, and were asked to try it out to demonstrate they could
Methods use the instrument. Throughout the study programme, the
participants were asked to record an exercise diary.
Study design and subjects
Primary and secondary outcomes
This was a prospective, single-centre, controlled, ran-
domised research study with 1:1 allocation to the control Outcome measures were recorded at baseline and at 12
group and the intervention group. Patients were identified weeks, with distance walked as the primary outcome mea-
from referrals to the vascular clinic at Cumberland Infirmary, sure. The data for claudication distance (i.e. the distance
Carlisle. Symptomatic peripheral arterial disease, Fontaine at which there is onset of claudication pain) and maxi-
IIa or higher, was diagnosed in patients by consultant vascular mum walking distance were obtained using a treadmill set
surgeons when first consulted on symptoms alone: exertional at 3.2 km/hour (as reported by Hiatt et al. [13]) and a 10◦
calf pain upon walking and an ankle brachial pressure index gradient. If no claudication pain developed, the maximum
<0.90 [3,11]. Adult patients with symptomatic intermittent walking distance was recorded as the claudication distance.
claudication due to peripheral arterial disease who were capa- The secondary outcome measure was the ankle brachial pres-
ble of giving informed consent were invited to participate sure index, which was measured with a handheld Doppler
in the study; there were no age restrictions. Patients were machine. This tool is commonly used for diagnosis and mon-
excluded according to the following criteria: unstable car- itoring of peripheral arterial disease [13]. In healthy persons,
dio/respiratory condition, such as uncontrolled hypertension, the ankle brachial pressure index is at least 1, with the sys-
cardiovascular accident or myocardial infarction within the tolic blood pressure equal in all limbs or higher in the ankle
last 2 months; surgery within 6 weeks of enrolment; major [3]. The highest measure of the dorsalis pedis and posterior
amputation of one or more lower limbs; and blood glucose tibial pressure was divided by the brachial pressure in the
level >13 mmol/l (i.e. uncontrolled diabetes) [10,12]. Follow- right arm of the patient. Since the recruited participants had
ing consent, the patients were allocated at random to either the claudication symptoms in either the right leg, left leg or both
control group or the intervention group. The study was per- legs, the ankle brachial pressure index score for either the
formed unblinded, and a non-restricted randomised sequence affected leg (unilateral disease) or the average of the ankle
246 N. Tebbutt et al. / Physiotherapy 97 (2011) 244–249

Table 1
Demographic overview of and completion rate in the two study groups.
Control group (n = 18) Intervention group (n = 24) P-Valuea
Mean age (range) 71 (47 to 86) 66 (45 to 77) 0.13
Gender (male/female) 13/5 15/9 0.52
Smoker 5 (28%) 7 (29%) 0.92
Bilateral/unilateral peripheral arterial disease 5 (28%)/13 (72%) 7 (29%)/17 (71%) 0.91
Diabetes 0 (0%) 2 (8%) 0.22
Beta-blockers 1 (6%) 2 (8%) 0.74
Statins 17 (94%) 19 (79%) 0.16
Completion rate 15/18 (83%) 20/24 (83%) 1
a Comparison of two study groups by one-way analysis of variance.

brachial pressure index scores for both legs (bilateral disease) intervention group but this difference was not statistically
is presented to account for this. significant.
The claudication distance and maximum walking distance
were measured for all 35 participants who completed the
Data analysis and sample size
study (15 in the control group and 20 in the intervention
group); however, the ankle brachial pressure index was not
In order to determine the required sample size, the assump-
available for one control participant at baseline and was there-
tion was made that distribution of data would be normal,
fore excluded from this outcome measurement. The trial was
and it was estimated – based on previous studies – that the
discontinued when an interim analysis was performed of the
mean baseline maximum walking distance for all participants
data. The primary reason was the variance in the values of the
would be 300 m with a standard deviation of 100 m [9,10].
primary outcome measures (claudication distance and max-
It was estimated, for the purpose of an a-priori sample size
imum walking distance, see results below). No unintended
calculation, that the intervention group would improve by
effects were observed in any of the participants. The data for
20% after 12 weeks, with the control group remaining static.
the 42 participants enrolled, rather than the 90 participants
Applying a two-tailed unpaired t-test to compare the means of
originally planned, are given below.
the control and intervention groups, and taking into account
80% power, 5% significance plus a 10% participant dropout
rate, the total number of participants required was 90 (45 in Claudication distance and maximum walking distance
each group). The changes in claudication distance, maximum
walking distance and ankle brachial pressure index between Both the claudication distance and the maximum walk-
baseline and 12 weeks in the control and intervention groups ing distance measured in all patients showed a wide range
were determined. Statistical analyses were performed using (Figs. 1 and 2, respectively). The claudication distance
Statistical Package for the Social Sciences Version 17 (SPSS achieved by patients varied enormously; for example, from
Inc., Chicago, IL, USA). 10 m to 520 m in the control group at baseline. The median
distance walked before the development of claudication
symptoms was 100 m [interquartile range (IQR) 40 to 137 m]
Results for the control group at baseline and remained at 100 m (IQR
60 to 180 m) at 12 weeks. For the intervention group, the
In total, 42 patients were recruited into the study; 18 median distance walked before the development of claudi-
in the control group and 24 in the intervention group. The cation symptoms changed from 65 m (IQR 50 to 110 m) to
CONSORT guidelines require a statement on the number of 85 m (IQR 53 to 138 m) over the 12-week period. It should
patients assessed for eligibility [14]. This was not recorded be noted that participants did not always stop walking due
in this study, therefore no data are available on the num- to claudication pain. One control participant and three inter-
ber of patients who did not meet the inclusion criteria or vention participants did not experience any claudication pain
who declined to participate. Of the 42 patients recruited, before finishing their walking test; shortness of breath and
three patients in the control group and four patients in the tiredness were the reasons on these occasions. Where no clau-
intervention group discontinued during the 12-week study dication pain was experienced, the claudication distance was
programme. The reasons were surgical intervention (one case considered to be equal to the maximum walking distance. The
in control group and one case in intervention group – both maximum distance walked by participants ranged from 40 m
had angioplasty) or no reason given (two cases in control for one patient in the control group at baseline to 770 m for
group and three cases in intervention group). The control another patient. Over the 12-week trial programme, the max-
group and the intervention group did not differ significantly imum distance walked by participants in the control group
in the demographic and clinical make-up of the participants worsened; the median distance reduced from 260 m (IQR
(see Table 1). More diabetic patients were allocated to the 100 to 325 m) to 210 m (IQR 140 to 430 m). However, for
 N. Tebbutt et al. / Physiotherapy 97 (2011) 244–249 247

Table 2
Comparison of participants’ response rates at 12 weeks.
No. of participants with improved distance
Control group Intervention group
Claudication distance 7/15 (47%) 13/20 (65%)
Maximum walking distance 9/15 (60%) 14/20 (70%)

Fig. 1. Distance to claudication pain (CP) for the control and intervention
groups. Outliers not included in box and whisker plot are denoted with an
asterisk or dot.

participants in the intervention group, the median maximum

distance walked improved from 160 m (IQR 103 to 274 m)
to 200 m (IQR 110 to 308 m) at the 12-week endpoint. Fig. 3. Ankle brachial pressure index (ABPI) score in the control and inter-
vention groups. Outliers not included in box and whisker plot are denoted
Maximum walking distance and claudication distance
with an asterisk or dot.
were further analysed to determine how many participants
actually showed an improvement in distance walked for these similar in both study groups, although more participants in
two parameters. Table 2 shows that more participants who the intervention group improved their claudication distance.
used the Step It pedal improved their claudication distance
compared with those in the control group. The number of par- Ankle brachial pressure index
ticipants who improved their maximum walking distance was
A secondary objective of this study was to conduct ankle
brachial pressure index measurements to determine whether
the plantar flexion exercise pedal could possibly have a pos-
itive effect. As shown in Fig. 3, the median ankle brachial
pressure index remained virtually unchanged over the 12-
week study period. A marginal improvement was seen in the
control group (baseline: 0.73, IQR 0.56 to 1.00; 12 weeks:
0.74, IQR 0.61 to 0.97), and the median ankle brachial pres-
sure index decreased by a fraction in the intervention group
(baseline: 0.64, IQR 0.51 to 0.87; 12 weeks: 0.63, IQR 0.54 to
0.73). Four control participants and three intervention partic-
ipants had an ankle brachial pressure index >0.9 at baseline,
despite the fact that patients had originally been diagnosed
with peripheral arterial disease by the vascular team using
ankle brachial pressure measurements as a diagnostic tool.

Discussion

The first clinical study involving the Step It pedal, con-

Fig. 2. Maximum walking distance (MWD) in the control and intervention ducted on healthy volunteers, was published recently [15].
groups. Outliers not included in box and whisker plot are denoted with a dot. The hypothesis that increased calf muscle exercise may alle-
248 N. Tebbutt et al. / Physiotherapy 97 (2011) 244–249

viate the claudication symptoms experienced in peripheral in terms of sample size (total n = 21), raising the possibility
arterial disease and thereby reduce the pain experienced in for obtaining a false-negative result. Another complication of
the lower extremities was therefore followed up using the using the ankle brachial pressure index as an outcome mea-
Step It pedal in this pilot study. The primary objective of this sure is that the measurements may differ depending on the
study was to determine whether an unsupervised exercise pro- experience of the person performing the measurements and
gramme involving a plantar flexion pedal that stimulates calf the method applied [18,19].
muscle exercise can potentially reduce claudication symp- In another pilot study, involving a total of 25 patients with
toms in patients with peripheral arterial disease. There was peripheral arterial disease, Wang et al. randomised partici-
little difference in the percentage of patients in the con- pants to either a control group or a plantar flexion exercise
trol group and the intervention group who improved their group and showed a 20% increase in time to exhaustion for
walking distance over the 12-week programme, as measured the plantar flexion group [20]. The plantar flexion device used
by claudication distance and maximum walking distance, on that occasion was an ergometer (a pedal attached to a exer-
and the effect size seen in either group is not likely to be cise bike), and the supervised training involved 4 × 4-minute
clinically relevant. The average percentage improvement in exercise intervals three times per week. Together with the
the median maximum walking distance for the intervention data obtained in the present study, it may be that localised
group was 25%, whereas an improvement of 30% is con- exercise of the calf muscles alone is insufficient to resolve
sidered clinically significant [16,17]. As mentioned above, claudication symptoms due to the limited resistance of the
the considerable variance in distance walked by patients pedal itself, limited exercise time or lack of whole-body aer-
led to the discontinuation of the study. Bearing in mind obic exercise. It may, however, be of use to prepare patients
that peripheral arterial disease is a progressive disorder before they embark on a more strenuous aerobic exercise pro-
that worsens over time, a positive note is that neither the gramme, although more research on a larger scale is needed
control nor the intervention group deteriorated very much [20,21].
over the measured 12-week period in terms of measurement Part of the reason for using the Step It pedal in the present
outcomes (Figs. 2 and 3). Other studies have shown that study was due to the notion that even very unfit people can
exercise programmes increase the fitness of patients with use it. Originally, the Step It pedal was designed for use in
peripheral arterial disease by 50% to 200% [5,9]. The com- aeroplanes for long haul flights, so that the lower leg could
pleted diaries provided by the participants did not indicate be exercised with the aim of reducing ‘economy syndrome’.
a high degree of non-compliance with the Step It exercise The data obtained with the present sample show a consid-
programme. erable degree of variance, and therefore it was not possible
In order to gain an understanding of why the patients in this to determine if a significant benefit is associated with use of
study did not improve dramatically, the baseline performance the Step It pedal. However, this could potentially be achieved
of participants in other studies was assessed. In studies by with a larger sample, provided that the study is better con-
Zwierska et al. and McDermott et al., the mean maximum dis- trolled in terms of patients’ Fontaine classification, and the
tance that participants could walk was approximately 300 m maximum and minimum distance that subjects are able to
(standard deviation approximately 90 m) [8,10]. A Cochrane walk at baseline.
report also reported a baseline maximum distance walked by This study has a number of limitations. Firstly, the sample
patients of 300 m [9]. In the present study, the median max- size was small. As this was a single-centre study, only patients
imum walking distance at baseline ranged from 160 m for from a certain area were recruited, which together with the
the intervention group to 260 m for the control group. This limited sample size means that the results may not be repre-
implies that a proportion of the patients in the sample were sentative of a wider population. Cumbria is a county in the UK
less fit than those in the two abovementioned studies which with a high prevalence of cardiovascular disease in its popu-
demonstrated a positive effect of an exercise programme. lation. The two study groups were, however, well balanced in
In the study by Zwierska et al., mean ankle brachial pres- terms of demographics and comorbidities, and this study does
sure index at baseline was 0.65 (improving to 0.68 after a add to the recent literature about the application of specific
24-week exercise programme), which is similar to the median lower limb exercise programmes in patients with peripheral
of 0.64 for the intervention group in the present study [8]. In arterial disease [7,8,10,13,20]. One other drawback may be
their study, the index increased to 0.68 after a 24-week exer- that the exercise programme may not have been sufficiently
cise programme, whereas the median ankle brachial pressure intensive to improve fitness levels. Nevertheless, a 20-minute
index in the intervention group in the present study decreased exercise session was likely to represent an increase compared
to 0.63. These results are in line with other evidence which with the patients’ usual level and frequency of exercise, and is
shows that the ankle brachial pressure index does not change in alignment with other plantar flexion exercise programmes
significantly with exercise in patients with peripheral arterial used [20]. With unsupervised exercise, there is always a risk
disease [6]. The present data are also concordant with one that participants do not comply with the programme. It has
other study that evaluated a year-long exercise programme been shown that those in supervised exercise programmes
in patients with peripheral arterial disease [16]. The study by fare better than patients who are asked to exercise unsuper-
Crowther et al. had a similar drawback to the present study vised [9].
 N. Tebbutt et al. / Physiotherapy 97 (2011) 244–249 249

In conclusion, although exercise has been shown to be symptomatic peripheral arterial disease: a randomized controlled trial.
beneficial in general in patients with peripheral arterial dis- J Vasc Surg 2005;42:1122–30.
[9] Bendermacher BLW, Willigendael EM, Teijink JAW, Prins MH. Super-
ease, further larger scale research is needed to determine
vised exercise therapy versus non-supervised exercise therapy for
whether intermittent claudication symptoms can be solved intermittent claudication. Cochrane Database Syst Rev 2006, Issue 2.
by applying exercise programmes that concentrate on the calf Art. No.: CD005263. doi:10.1002/14651858.CD005263.pub2.
muscles. The pilot data obtained in this study could inform [10] McDermott MM, Ades P, Guralnik JM, Dyer A, Ferrucci L, Liu K, et
researchers when devising a research protocol and determin- al. Treadmill exercise and resistance training in patients with periph-
ing the required sample size. eral arterial disease with and without intermittent claudication. JAMA
2009;301:165–74.
Ethical approval: National Health Service National Research [11] Scottish Intercollegiate Guidelines Network. Diagnosis and manage-
Ethics Service, West Midlands Committee (Ref. No. ment of peripheral arterial disease. Guideline no. 86. Edinburgh: SIGN;
2006.
09/H1014/38).
[12] Chipkin SR, Klugh SA, Chasan-Taber L. Exercise and diabetes. Cardiol
Funding: The Step It pedals were provided by the Step It Clin 2001;19:489–505.
company. [13] Hiatt WR, Regensteiner JG, Hargarten ME, Wlfel EE, Brass EP. Benefit
of exercise conditioning for patients with peripheral arterial disease.
Conflict of interest: None declared. Circulation 1990;81:602–9.
[14] Schulz KF, Altman DG, Moher D, CONSORT group. CONSORT 2010
statement: updated guidelines for reporting parallel group randomised
trials. BMJ 2010;340:c332.
References [15] Bergqvist D, Landin G, FLynner K, Heide B, Kusoffsky A. Anti-
swelling effect of Step It—an active foot pump system. J Am Podiatr
[1] Weitz JI, Byrne J, Clagett GP, Farkouh ME, Porter JM, Sackett Med Assoc 2009;99:20–2.
DL, et al. Diagnosis and treatment of chronic arterial insufficiency [16] Crowther RG, Spinks WL, Leicht AS, Sangla K, Quigley F, Golledge J.
of the lower extremities: a critical review. Circulation 1996;94: The influence of a long term exercise program on lower limb movement
3026–49. variability and walking performance in patients with peripheral arterial
[2] Milani RV, Lavie CJ. The role of exercise training in peripheral artery disease. Hum Mov Sci 2009;28:494–503.
disease. Vasc Med 2007;12:351–8. [17] Hiatt WR, Regensteiner JG, Hargarten ME, Wlfel EE. Benefit of
[3] Dieter RS, Chu WW, Pacanowski JP, McBride PE, Tanke TE. The exercise conditioning for patients with peripheral arterial disease. Cir-
significance of lower extremity peripheral arterial disease. Clin Cardiol culation 1990;81:602–9.
2002;25:3–10. [18] Kaiser V, Kester ADM, Stoffers H, Kitslaar P, Knottnerus JA. The
[4] Creasy TS, Fletcher EWL, Collin J. Peripheral vascular disease. BMJ influence of experience on the reproducibility of the ankle-brachial
1989;299:1166. systolic pressure ratio in peripheral arterial occlusive disease. Eur J
[5] Fowkes FGR, Housley E, Cawood EHH, MacIntyre CCA, Ruckley CV, Vasc Endovasc Surg 1999;18:25–9.
Prescott RJ. Edinburgh Artery Study: prevalence of asymptomatic and [19] Jeelani NU, Braithwaite BD, Tomlin C, MacSweeney ST. Variation
symptomatic peripheral arterial disease in the general population. Int J of method for measurement of brachial artery pressure significantly
Epidemiol 1991;20:384–92. affects ankle-brachial pressure index values. Eur J Vasc Endovasc Surg
[6] Watson L, Ellis B, Leng GC. Exercise for intermittent claudication. 2000;20:25–8.
Cochrane Database Syst Rev 2008;4. CD000990. [20] Wang E, Hoff J, Loe H, Kaehler N, Helgerud J. Plantar flexion: an
[7] McGregor GS, Whyte G, Buckley JP, Houghton. The effect of weight- effective training for peripheral arterial disease. Eur J Appl Physiol
bearing and non-weight-bearing exercise on patients with intermittent 2008;104:749–56.
claudication. J Sports Sci 2000;19:59–60. [21] Helgerud J, Wang E, Mosti MP, Wiggen ON, Hoff J. Plantar flexion
[8] Zwierska I, Walker RD, Choksy SA, Male JS, Pockley AG, Saxton JM. training primes peripheral arterial disease patients for improvements in
Upper- vs lower-limb aerobic exercise rehabilitation in patients with cardiac function. Eur J Physiol 2009;106:207–15.

Available online at www.sciencedirect.com