Therapeutic Effectiveness of Electric Stimulation of the Upper-Limb Poststroke Using Implanted Microstimulators

Participants 

A convenience sample of subjects was recruited from the community through advertisements in stroke clubs and newsletters. The criteria for inclusion were adults over 18 years who were more than 6 months poststroke, with a hemiparesis resulting in weakness of wrist and elbow extension, but with perceivable voluntary wrist extension up to 75% of passive range. All subjects had sufficient volitional finger flexion to grip, elbow flexion to bring the hand to the mouth, and shoulder flexion to lift the arm away from the body against gravity to an angle of approximately 45°. Subjects with severe spasticity (MAS score >3) were excluded because it has been found previously that stimulation to open a spastic paretic hand during volitional movement can be problematic.18 Subjects were screened medically, and those with any active device implant (eg, pacemaker), uncontrolled epilepsy, or history of endocarditis, prosthetic valve, or murmur were excluded. Because this study was investigational and demanding on subjects' time, psychologic status (Hospital Anxiety and Depression Scale), cognitive status through the Mini-Mental State Examination,19 and their attitudes and beliefs toward their condition and rehabilitation through the Multidimensional Health Locus of Control Scales20 were assessed. Ethics committee approval (no. 04/12/021) and Medicines and Healthcare products Regulatory Agency (no. CI/2004/0027) approval was granted. Subjects gave written informed consent, and the study was performed according to institutional guidelines for good clinical practice in research.

Radiofrequency Microstimulator System 

Radiofrequency microstimulators (fig 1) are miniature, single-channel, implantable stimulators that produce capacitively coupled, charge-balanced, asymmetric, biphasic, constant-current pulses. Radiofrequency microstimulators receive power supply and control signals from an external control unit through an inductance coil that, in this application, was worn as 2 cuffs around the arm and forearm (fig 2). A choice of 5 sizes of both arm and forearm cuffs were available and ensured a snug fit and ease in donning and doffing, which all subjects could do without help. Each device was individually addressed, allowing control over timing, ramp rise and fall times, current amplitude, and pulse width. These stimulation parameters were programmed through a personal computer−based fitting system and downloaded to each person's control unit. The control unit measured 35×75×140mm and weighed 400g. While performing their exercise regimen, subjects placed it either on a table in front of them or on another surface nearby.

View Large Image Download to PowerPoint

Fig 1. The radiofrequency microstimulator (actual size, 17×2.4mm) showing the eyelet used for attachment on the suture.

View Large Image Download to PowerPoint

Fig 2. The external components of the radiofrequency microstimulator system.

Implantation Procedure 

By using a minimally invasive day-case procedure,21 implantation was performed by using custom-designed implantation tools in a sterile operating room under local anesthesia. Postoperative paracetamol was prescribed for pain management. One week before implantation, optimal electrode positions had been identified by using needle electromyography electrodes. Coordinates were documented and insertion sites marked on the skin. Using these as a guide, for each device a 5-mm skin incision was made. A stimulation probe was used to confirm that the proposed implantation site elicited the desired functional movement, and a dilator with sheath was then introduced over the probe. The probe was then removed; the radiofrequency microstimulator device was inserted through the sheath and ejected into the tissues by using an ejection tool. At each stage of the process, implantation site and thus desired movement was confirmed by using either the stimulation probe (preinsertion of the device) or an external coil to activate the radiofrequency microstimulator device. If the activation of the device did not elicit the desired movement, it was removed by using a suture attached to an eyelet on the device and reinserted. For each subject, 2 radiofrequency microstimulator devices were implanted close to the radial nerve branches to the medial and lateral heads of the triceps to activate elbow extension. To extend the fingers and extend and abduct the thumb for hand opening, either 1 or 2 devices were positioned close to the PIN. Wrist extension was activated by either 2 or 3 devices close to the motor points of ECU and ECRL.

Programming and Home-Exercise Regimen 

Design and programming of stimulation sequences for the home-exercise regimen was undertaken by a research physical therapist and began at 2 weeks postimplantation. At the initial fitting session, both the threshold and upper limit, defined as the level above which stimulation became (1) uncomfortable, (2) response reached a plateau, or (3) there was overflow to other muscles, were set for each microstimulator. An algorithm for controlling stimulation output provided stepwise increases in both pulse width and current amplitude so that the charge was minimized. Levels were checked and adjusted if necessary at weekly intervals during the first 4 weeks and at 4 weekly intervals thereafter. Thresholds remained stable.

Specific functional activities were identified for each subject relating to their goals and ability level. Activity programs were designed for each functional activity and were defined by output and timing of stimulation, including rise and fall times, for each microstimulator singly and, when the functional activity required 2 or more devices to be active in synergy, for the combined output so that the movement pattern was optimized. Activity sequences were programmed by using an intuitive, drag-and-drop graphics interface on a personal computer, which allowed rapid adjustment of stimulation parameters. Up to 3 different activity programs were then downloaded and stored in the control unit. The choice of program and level of stimulation (proportional increase on all channels on the selected program) were controlled by the subject by using easy-to-grasp knobs on the control unit.

Subjects attended the laboratory twice weekly during the first 1 to 3 weeks for programs to be designed and to learn how to use the system. As soon as subjects were able to use the system independently and perform the chosen functional activities, they entered the 12-week home-use period. During this period, they continued to attend the laboratory either weekly or on alternate weeks for approximately 1-hour sessions so that adjustments could be made by the research therapist to the stimulation programs and activities progressed in terms of speed at which they were performed and from simple to more complex tasks. Lower-functioning subjects and those with high levels of spasticity were initially set up with simple cyclic stretching exercises alone or before simple functional tasks, but emphasis was on using stimulation to augment voluntary task-based movement. Examples of tasks were opening the hand to grasp a can with the arm supported on a table (coordinated finger and wrist extension but without elbow extension) or polishing a table (elbow extension without hand opening) and, when function was limited by weakness, strengthening exercises such as extending the elbow against a resistance using Therabands.b Subjects who already had, or progressed to having, sufficient voluntary activity without excessive finger spasticity used stimulation programs to practice more complex functional reach-and-grasp tasks (fig 3). These involved coordinating multiple joint movements such as reaching (activation of elbow extensors), opening of the hand (adding wrist and finger extension), and, by switching off stimulation to the PIN and thus releasing finger extension, voluntary grasping of a variety of everyday objects while the grip was enhanced by maintained activation of the wrist extensors. Switching off elbow extensors enabled the object to be brought toward the body so that it could be used (eg, drinking from a cup, using a knife to cut vegetables, brushing hair). After use, the activation of elbow extension and then the PIN allowed the object to be replaced and released. During the 12-week home-exercise period, subjects were asked to practice at home using everyday objects for at least 1 hour a day 5 days a week. Subjects were encouraged to practice tasks without using stimulation especially immediately after an exercise session. An automated activity log in each subject's control unit provided a record of use of the stimulator.

View Large Image Download to PowerPoint

Fig 3. An example of an activity sequence to support reaching, grasping, and releasing an object.

Outcome Measures 

Outcome measures were applied by the same assessor: 4 weeks and 1 day before implantation (baseline 1 and 2) and at the end of the 12-week home-exercise period, which was between 14 and 17 weeks postimplantation. The first baseline assessment familiarized subjects with the testing procedures and was not used in the analysis. The primary outcome measure for upper-limb function was the ARAT,22, 23 and the primary impairment measure was the ability to perform a tracking test (motor control). The ARAT is a standardized measurement tool that evaluates 4 domains of upper-limb function at an activity level: grasp, grip, pinch, and gross motor function. The total score is the sum of the subscores from the 4 domains, ranging between 0 and 57, with higher scores indicating better function. This measure is widely used in electrical stimulation and other upper-limb rehabilitation studies.11, 24, 25 Reliability and validity26, 27, 28 have been established, with a 5.7-point change in score suggested as being clinically important.29 The ARAT was recorded on video for repeat scoring by a moderator who was blinded to which stage of the trial the video assessment related.

An instrumented wrist rig30 was used to quantify the following impairments: motor control by accuracy during a tracking task (tracking index), spasticity by the surface electromyographic response to passive stretch of wrist flexors at 1.5Hz termed stretch index, AROM, and isometric force (flexion, extension). Before the study, the wrist rig had been tested for reliability and validity with a group of patients with stroke,31, 32, 33, 34 and ranges of values for an unimpaired age-matched population were identified. Test-retest reliability was shown to be excellent, with intraclass correlation coefficient values ranging from .86 to .98.34 Repeatability for each measurement using the wrist rig was defined by the repeatability coefficient,34 which is the size of change required to be reasonably certain (95%) of a real difference not caused by measurement error. By using this range value for each measurement, we were able to detect real change in each subject's performance.

For all wrist rig measures, the subject sat in a specially designed chair with the arm to be tested secured to a molded splint lined with an inflatable pouch that held the hand in midposition (pronation, supination) with the thumb uppermost and the fingers and thumb in comfortable extension. The arm and forearm were secured by 3 self-adhesive (Velcro) straps. The arm support was adjusted so that the shoulders were neither depressed nor elevated, with the elbow at 90° of flexion and the fulcrum aligned with the wrist joint. The rig allowed virtually friction-free movement of the wrist in the horizontal plane.

The raw analog angle and force data were recorded in real time at a sample rate of 1KHz and stored for offline processing by using LabView.c Data analysis to generate metrics and indices was performed by using software developed in the Matlabd environment. Surface electromyographic signals were recorded by using standardized electrode positions and skin preparation.35 Electromyographic signals were recorded by using BEAC Biomedical preamplifierse connected to self-adhesive electrodes.f Signals were amplified and antialias filtered by using a low-pass filter (cutoff frequency, 250Hz).

For the tracking test, a screen (computer monitor) in front of the subject displayed an elliptical target (target signal) moving sinusoidally horizontally across the screen at a frequency of 0.5Hz through 40° (20° either side of midpoint). The angular position of the wrist (tracking signal) was displayed as a cross, allowing the subject to visually track the target by flexing and extending the wrist trying to keep the cross within the ellipse. Subjects practiced the test several times until they believed they had achieved their best performance. They then rested for 3 to 5 minutes before performing the test for 60 seconds, during which position and electromyographic data were acquired. Accuracy in performance of the tracking task (motor control) was defined as the cross-correlation between the target and tracking signals (tracking index).32 To measure the stretch index, the same target signal was displayed but at a frequency of 1.5Hz. The subject was asked to relax and avoid any contribution, either assistive or resistive to the procedure, while the assessor moved the wrist through 60° about the midpoint maintaining phase and displacement with the target signal for 40 seconds. Position and electromyographic data were acquired. The electromyographic response to rapid passive wrist extension was quantified by the stretch index, which is calculated from the root mean square electromyographic activity that exceeds baseline resting values during the period between 0° (midpoint) and peak extension and is expressed as millivolts. AROM (in degrees) and peak isometric force (in newtons) were also measured in the rig. For the isometric force test, the wrist was positioned at midpoint and immobilized. The subject was instructed and encouraged verbally to exert maximum force first into extension and then flexion. Force was maintained for 5 seconds, and the maximum value was used in the analysis.

The movement, coordination, and sensation of the upper limb were measured by using FMA upper limb, a valid and reliable measure of poststroke impairments.27 The resistance to passive movement was measured by the MAS.36 The order in which tests were conducted was the same for each subject and at each assessment. Wrist rig measures were conducted in the order as follows: AROM, stretch index test, tracking test, and isometric force.

Statistical Analysis 

Statistical analysis was performed by using SPSS.g Means ± SDs of all outcome measures at baseline 2 and 12 weeks are presented. Mean changes ± SD are reported and statistically significant differences between baseline 2 and 12 weeks have been estimated by using the Wilcoxon signed-rank test. We derived the 95% CI from the standard error of the mean change generated by a paired t test. Individual changes in all outcome measures are presented and we have identified where the size of change in ARAT and FMA was clinically important, as suggested in a recent study,26 and for wrist rig measurements where the effect size was considered a real change as opposed to measurement error based on the previously shown repeatability coefficients. A change of 1 point on the MAS is accepted as clinically important.

Personalized Activity Programs 

Activity programs in which voluntary movement was augmented by appropriate stimulation were designed to assist each subject in achieving his/her specific goals and tasks. Goals were only considered to have been achieved when the subject could perform the task without stimulation. Subject 1, who was 10.5 years poststroke, had a baseline ARAT score of 39, limited forward reach, and inability to release objects. The goals were to grasp and release objects away from the body and catch a ball with both hands. To achieve these goals, activity programs progressed from elbow-extension activities, to hand opening with the arm supported, and then to unsupported elbow extension and hand opening to grasp, use, and replace an object such as a glass of drink with the affected hand or a ball using both hands. Initially, she could open her fingers to grasp and release objects more effectively and with a more normalized movement pattern with stimulation. At the end of the 12-week period, both goals were achieved with a satisfactory but slow movement pattern, although she still could not achieve full elbow extension.

Subject 2, who was 16 months poststroke, had a baseline ARAT score of 8, was unable to open his hand, and had limited forward reach, all of which were associated with spasticity. His main goal was to be able to grasp and hold an object such as a can of drink so that it could be opened with the nonaffected hand. Activity programs progressed from single-joint extension at the elbow, wrist, and fingers, to elbow extension while weight bearing through the hand, and finally to reaching and grasping with the arm supported on a table to allow effective hand opening. At the start of the study, even with stimulation, he was unable to perform functional tasks, but, by the end of the 12-week period, he was able to grasp and release small objects close to his body without simulation. With stimulation, he was able to grasp and release larger objects further away from his body. The goal was achieved, and spasticity was reduced. Functional improvement was sufficient to enable him to swing a golf club and hit a ball, which he had not done since his stroke.

Subject 3, who was 13 months poststroke, had a baseline ARAT score of 16 and particular difficulty in opening the hand when the elbow was extended. His goals were to grasp a door handle and open a door, to perform bilateral reach and grasp tasks, and to eat with a knife and fork. Exercises augmented by stimulation progressed from single-joint extension activities to reaching to touch objects, bilateral reaching to grasp a bowl, elbow extension against a resistance, and finally to use the system to reach and grasp a jug and pour water from it. At the end of the 12-week period, he was able to grasp a door handle and eat with a knife and fork. Both tasks could be performed equally well with or without stimulation. Movement pattern was good but slightly slower and less well controlled than normal.

Subject 4, who was 2 years and 3 months poststroke, had a baseline ARAT score of 28, a weak grasp, and a poor release, especially of larger objects. Tremor on effort, which interfered with function, was identified as a specific problem. Specific goals were to be able to wash his body with the affected hand and to reach, grasp, and release without tremor. This subject put considerable effort into activities that progressed from supported to unsupported arm, wrist, and finger extension, and bilateral and unilateral activities reaching to different positions, some above the horizontal. During the first 6 weeks of the exercise program, the strength of contraction elicited by the devices in the triceps (medial heads of triceps, lateral heads of triceps) failed to generate a sufficiently strong contraction to provide functional elbow extension. After the 12-week assessment, 2 further devices were implanted that have worked satisfactorily. Because of poor elbow extension, especially against gravity, this subject had difficulty practicing activities involving high reaching, which may have explained why, although he had an above average improvement on the tracking task (102.8), which measured motor control at the wrist, he had no improvement in function (ARAT score decrease, 3 points). Tremor remained a problem; however, he was able to wash his body and hair by using both hands. Throughout the period, quality of movement was better with stimulation, and he may therefore have benefited from using the system as a neuroprosthesis.

Subject 5, who was just over 7 years poststroke, had a baseline ARAT score of 33 and some ability to grasp and release small- and medium-sized objects placed close to the body, although this deteriorated quickly with repetition. Her specific goals were to be able to grasp the handlebar brake when riding a bicycle, reach objects out of cupboards, and use a vacuum cleaner. Activities progressed from single-joint activities of the elbow, wrist, and fingers to reaching for a variety of objects at different heights and reaching, grasping, drinking, and replacing a drinking mug. The ability to use a vacuum cleaner with the affected hand was achieved, but she continued to have difficulty releasing objects when combined with the effort of reaching away from the body. Immediately after a period of cyclic stimulation that elicited the movement pattern required for a functional task (hand opening), she was able to perform the task more effectively and with a more normative movement pattern. Performance was always better with stimulation, even at the end of the 12-week period.

Subject 6, who was 19 months poststroke, had a baseline ARAT score of 30, was able to grasp small objects with reasonable dexterity, but was unable to release them; shoulder and elbow control was good. His goals were to pick up and carry moderately heavy large objects and to reach for objects in high cupboards. It was possible to design activities in which these could be practiced and performance of these tasks improved so that by the end of the 12 weeks he was able to release objects placed at high levels, although slowly and with greater than normative effort. Because of unrelated health problems, this subject only used the system for 36 days.

Subject 7, who was 3.2 years poststroke, had the lowest baseline ARAT score of 7 and had only a flicker of activity in the finger and wrist extensors, weak grasp, and inability to release objects. Goals were to hold objects such as a fork in the affected hand and carry a tray with both hands. Activity programs were designed for table polishing, elbow extension with the forearm supported, and this was progressed to grasp-and-release activities but still with elbow support. At the end of the study, this subject managed to carry a tray with both hands, but there was no further improvement in reach and grasp.

In summary, personalized functional exercise with programmed radiofrequency microstimulation resulted in a mean improvement in all outcome measures. Functional improvement (ARAT) was measured in 4 of the 7 subjects, 3 of whom made clinically important changes. Of those who improved functionally, 2 subjects (subjects 2, 3) also had a clinically important change in motor control, and improvement for subject 6 was to within the normative range. One subject (subject 2) who had the highest baseline spasticity score had a meaningful reduction in spasticity measured by the stretch index and a clinically important reduction in resistance to passive movement measured by the MAS. All subjects except subject 2, who ranked second highest on this measure at baseline, had an increase in AROM. Isometric extensor muscle force improved in 6 of 7 cases, and improvement was clinically important in subjects 5, 6, and 7. The 2 subjects who did not change functionally did improve in motor control (subject 4) and extensor muscle isometric force (subject 5).

Future Work 

Previous studies suggest that voluntary activation of stimulation is important in recovery. De Kroon et al14 in a systematic review that identified a positive effect of stimulation in 13 of 22 patient groups reported that positive results were more common when stimulation was triggered by voluntary movement. At the end of this first phase of the study, all subjects elected to take part in a second phase in which external touch and angle sensors will be used to trigger activities and provide temporal control over stimulation of different muscle groups. This approach, supported by De Kroon's finding, will enable activities to be performed in response to voluntary command and at the subject's preferred speed. On completion of the second phase, subjects will be able to continue to use the system for a further 5 years; this will enable us to examine the feasibility and acceptability of prolonged use of an implanted system. The longest duration of stimulation found in the literature was 3 months, reported by Sonde et al.39 Clinically, however, stimulation is often continued for considerably longer but with less good compliance.

Study Limitations 

One disadvantage of the current system is that because the devices do not have their own power supply an inductance coil needs to be worn around the arm (see fig 2), which anecdotally subjects reported being an encumbrance. In the future, battery-powered devices may become available and thus remove the need for the coil except for recharging.