H.J.P Fokkenrood

Innovative strategies for
intermittent claudication

towards a stepped care approach and new outcome measures




 Patients with IC (PAD stage 2-3 according to the Rutherford classification) and visiting the vascular outpatient clinic of our hospital between August and November 2012 were eligible for this study. This study was conducted with approval of the local medical ethics committee of the Catharina Hospital (Eindhoven, the Netherlands).

 Inclusion criteria

 Inclusion criteria were >3 months symptoms of IC, and an <0.9 ABI at rest or a fall in systolic ankle pressure by more than 20% after treadmill testing. A treadmill protocol with a fixed inclination of 8% at 3.2 km/h for a maximum of 5 min was used.

 Exclusion criteria

 Patients with walking difficulties other than due to IC were excluded (e.g. prior amputation, severe arthritis, COPD GOLD 3-4, congestive heart failure (>NYHA class II) as was the use of walking aids. Patients with recent (less than 12 months) vascular surgical intervention prior to the study were also excluded as were patients unable to understand all specifics of the study protocol or having insufficient knowledge of the Dutch language.


Video Observation and activity monitoring

 Specificsof patient's medical and surgical history were obtained followed by a physical examination and a check of inclusion and exclusion criteria. After signing an informed consent, a DP (DynaPort, McRoberts BV, The Hague, the Netherlands) attached to a neoprene belt was strapped around the patient's waist at the level of the midlower back (figure 1). The patient's hospital visit (e.g. waiting room, doctor's visit, vascular laboratory assessments, treadmill testing, etc.) was then continuously video-recorded (GZ-HM335BE, JVC, Yokohama, Japan). Consequently, patients were asked to walk the hospitals car parking lot as abnormal walking due to IC would possibly occur during this effort. Subjects were instruct to act and move as they would normally do. Patients were filmed anonymously. Two observers were randomly assigned to perform all video recordings. Video recording of the activities was considered as 'gold standard'.


Categorizing movements by video

 Table I depicts seven standard categories associated with daily activities including lying, sitting, standing, shuffling, locomotion, (device) 'not worn', and 'activity not recorded' (private actions such as visit to restroom). Specifics of each category and transitions between the seven categories were described in detail. A concise description

of categorical and transitional activities formed the basis for a subsequent evaluation of all video recordings. All recorded activities were scored in time per activity (in seconds) using annotation software (ELAN 4.4.0, Max Planck Institute for Psycholinguistics, Nijmegen, the Netherlands) and exported as Microsoft Excel files. Additionally, the number of steps of all walking activities per patient were counted. All video recordings were scored in duplicate by two observers allowing inter-rater reliability (IRR) verification.


DynaPort MoveMonitor (DP)

 We chose to use the tri-axial DP accelerometer (84 × 50 × 8mm, 70 g) to monitor activities. Compared to other activity monitors, the DP showed high correlations between indirect calorimetry and generated MET output whereas walking speed was correctly measured in a population with chronic obstructive pulmonary disease (19). The device consists of a tri-orthogonal orientated piezo-capacitive acceleration sensor, a rechargeable battery and removable SD card to store the acceleration data. A DP stores digital data for a maximum of seven days. The raw acceleration data lend itself to a pattern recognition approach using logical algorithms (MoveMonitor analysis software, version 2.6) for the classification of postures (lying, sitting and standing) and motions (locomotion and shuffling). The detection algorithm consists of 5 major parts as published previously (11) (12). The first step is gait period detection based on an intensity threshold. These potential gait periods are scanned using frequency analysis and a validated step detection method resulting in 3 categories: walking, active (but not walking), and static periods. Second, transition detection is performed to identify upward or downward transitions. The result is the identification of either up (standing) or down (lying or sitting). Subsequently, angle calculation based on sensor tilt is used to determine whether the down part of this vector can be identified as lying (<30°) or sitting. Next, shuffling separation divides the active (not walking) parts into 2 categories: shuffling and transitions. Shuffling is defined as all movement from A to B that is not walking. Thus, if the number of steps is less than 3 or the intensity and direction of the motion do not comply with the characteristics of walking, the movements are classified as shuffling. The results of the software analysis were returned in comma separated value (CSV) files. The reports listed six of the described activities per second (table 1), except for the "activity not recorded" category. Data obtained from the DP were synchronized with data generated by video recording (ELAN 4.4.0, Max Planck Institute for Psycholinguistics, Nijmegen, the Netherlands).

Data analysis

 The first analysis of data was aimed at determining inter-rater reliability (IRR) of video recording itself, considered as the gold standard. The duration of activity (in 0.1 seconds) per category was summed for each patient. These values were analyzed for skewness and kurtosis to obtain an impression of the distribution. In case of normally distributed data, IRR of video recording were determined by comparing total duration per activity between both observations using intraclass correlation coefficients (ICC) with a two-way mixed model and absolute agreement.

 Aim of the second analysis was to validate the DP by using two types of analyses. Transformation in an "activity per second" format was required for annotation files. For this purpose, a computer program was developed (MATLAB 7.14, MathWorks, Massachusetts). The software converted the annotated duration of activities into an "activity per second" format. Activities defined by the DP and by video recordings were than compared by matching each activity per second for the six overlapping categories (except for the "not recorded" category). The agreement between the DP and the gold standard was calculated per subject by adding up the duration that the activity codes matched and was expressed as a percentage of the total duration that an activity was observed on video. Non-agreement percentages per subject were defined as: (total duration that the video observation and the DP corresponded at the same moment for not 'activity category' / total duration that not 'activity category' was observed on video) x 100%. Sensitivity, specificity and predictive values were calculated by taking the mean of agreement or non-agreement values respectively for each activity category as suggested by Dijkstra et al. (11). The second type of determining the validity of the DP was using ICCs for step count. Gait characteristics as observed on video were again considered as 'gold standard'. Steps performed on the treadmill (to obtain ankle brachial indexes) were analyzed separately from walking during a patients' hospital visit. "Not recorded" periods >5 seconds were excluded from analysis. Outliers of agreement were estimated per activity category and defined as four times the standard deviation (SD) of agreement. ICC were considered strong if = 0.7, moderate between 0.3 and 0.7, and weak = 0.3. P-values < 0.05 were considered statistically significant. Statistical analysis was performed using SPSS Statistics (MAC OS X version 20.0).