TSR 13:2, Spring 2006

Contents

THOMAS LAND PUBLISHERS, INC.

vii Foreword

viii Information for Authors

x Letters to the Editors

1 Ottawa Panel Evidence-Based Clinical Practice Guidelines for Post-StrokeRehabilitation

3 Method4 Target population4 Literature search5 Study inclusion/exclusion criteria5 Post-stroke rehabilitation interventions7 Outcomes7 Statistical analysis9 Reviewing the guidelines9 Results of literature search

10 Results10 Clinical practice guidelines for therapeutic exercises

14 Summary of trials14 Efficacy20 Strength of published evidence compared with other guidelines20 Clinical recommendations compared with other guidelines

21 Clinical practice guidelines for task-oriented training24 Summary of trials24 Efficacy27 Strength of the published evidence compared with other guidelines27 Clinical recommendations compared with other guidelines

28 Clinical practice guidelines for biofeedback30 Summary of trials30 Efficacy32 Strength of published evidence compared with other guidelines33 Clinical recommendations compared with other guidelines

34 Clinical practice guidelines for gait training36 Summary of trials37 Efficacy40 Strength of published evidence compared to other guidelines40 Clinical recommendations compared with other guidelines

41 Clinical practice guidelines for balance training43 Summary of trials43 Efficacy45 Strength of the published evidence compared with other guidelines45 Clinical recommendations compared with other guidelines

45 Clinical practice guidelines for sensory interventions46 Summary of trials46 Efficacy48 Strength of the published evidence compared with other guidelines48 Clinical recommendations compared with other guidelines

48 Clinical practice guidelines for constraint-induced movement therapy49 Summary of trials49 Efficacy50 Strength of published evidence compared with other guidelines50 Clinical recommendations compared with other guidelines

51 Clinical practice guidelines for shoulder subluxation52 Summary of trials52 Efficacy53 Strength of published evidence compared with other guidelines54 Clinical recommendations compared with other guidelines

54 Clinical practice guidelines for electrical stimulation56 Summary of trials57 Efficacy59 Strength of published evidence compared with other guidelines60 Clinical recommendations compared with other guidelines

61 Clinical practice guidelines for transcutaneous electrical nerve stimulation (TENS)61 Summary of trials62 Efficacy63 Strength of published evidence compared with other guidelines63 Clinical recommendations compared with other guidelines

64 Clinical practice guidelines for therapeutic ultrasound64 Summary of trials64 Efficacy64 Strength of the published evidence compared with other guidelines64 Clinical recommendations compared with other guidelines

65 Clinical practice guidelines for acupuncture65 Summary of trials66 Efficacy67 Strength of the published evidence compared with other guidelines67 Clinical recommendations compared with other guidelines

68 Clinical practice guidelines for intensity and organization of rehabilitation74 Summary of trials75 Efficacy83 Strength of published evidence compared with other guidelines83 Clinical recommendations compared with other guidelines

84 Discussion86 Therapeutic exercises87 Task-oriented training88 Biofeedback88 Gait training91 Balance training91 Sensory interventions

v

91 Constraint-induced movement therapy92 Shoulder subluxation93 Electrical stimulation94 TENS94 Therapeutic ultrasound95 Acupuncture95 Intensity and organization of rehabilitation

96 Conclusion

97 Acknowledgments

97 References

119 Appendixes

vi

All figures and tables cited in this issue are included in the CD attached to the insideback cover. This CD also includes an electronic version of the full text of this issuewith embedded hyperlinks to the figures and tables.

Additional copies can be ordered at www.thomasland.com

Foreword

vii

The concept of developing and implementing clinicalpractice guidelines has received considerable attentionin recent years among many medical professionals, clini-cal leaders, payers, and policy makers. It is interesting tonote that adoption of these tools has been considerablyless enthusiastic among rehabilitation programs andpractitioners than among many other disciplines andprofessionals. Reasons for the slower utilization of guide-lines are many, including concern that they might limitthe important role of creativity and ingenuity in theclinical practice of rehabilitation, relative deficiency ofsufficient evidence supporting many prevailing rehabili-tation practices, and an apparent lack of compelling rea-sons for clinicians to examine the possibility of changingexisting practices. There also is limited interest in thedevelopment and use of guidelines among rehabilita-tion professionals and a general lack of awareness oftheir usefulness among clinicians and their leaders. Al-though there are some notable prominent exceptions,clinical practice guidelines are in their infancy in rehabili-tation.

There is a school of thought that asserts that the expe-rience of developing these guidelines is as importantand beneficial as the actual implementation of thesetools. The opportunity to thoroughly and critically re-view existing medical literature and assess its implica-tions for clinical practice often is a unique and favorablecollaborative experience for clinicians and scholars.

However, it is in their roles in education and in pro-moting both quality and consistency of care that clinicalpractice guidelines demonstrate their greatest benefits.For trainees and for junior clinicians with minimal practi-cal experience, these tools can be used as an effectivetraining device. For all practitioners, they can be used asa template to ensure that favorable evidence-drivenpractices are being implemented. This is a method toimprove quality of care. These documents are tools and,like other instruments used by clinicians, they serve tosupport and facilitate successful rehabilitation; they donot replace the role of originality and they do not reducethe value of hands-on, interactive, collaborative prob-lem-solving by clinicians and patients that often is thehallmark of team-driven clinical stroke rehabilitation.

Topics in Stroke Rehabilitation recently has publishedseveral Clinical Practice Guidelines (TSR 10:1 and 10:2[2003], 11:4 [2004]). The frequency with which thesedocuments are being published at this time reflects theemerging interest among thought leaders and practitio-

ners in the performance and dissemination of compre-hensive reviews of literature and in the implementationof clinical practices that are based on strong evidence,where it exists.

In the present issue, Dr. Lucie Brosseau and her col-leagues in Ottawa, Ontario, Canada, have conducted anexhaustive review of the literature, employed rigorousgrading measures to rate the quality of the researchstudies, developed and applied a systematic method fororganizing key points from the manuscripts that theyreviewed, and created an extensive set of guidelines forclinical stroke rehabilitation practices that derive directlyfrom these findings. The product focuses on 147 specificrecommendations concerning 13 rehabilitation inter-ventions. Practicing according to these guidelines canbe expected to enhance the quality of care that is pro-vided by clinicians and to improve the level of function-ing that is experienced by patients.

Dr. Brosseau and her team are to be congratulated forthe enormous effort that was put into the production ofthis document, not only for their rigorous reviews andclear writing, but also because of the way in which theinformation is presented. Because the volume of mate-rial is potentially overwhelming, it could be somewhatunwieldy for most practitioners. This is one reason thatthe organization of the material is so relevant in order tofacilitate understanding and use of the recommendationsand their foundation. The managing editor, Mary Killion,also should be recognized for her extensive contributionto the layout and formatting of the data. The use of theCD-ROM is a novel approach. Displaying information in atabular format is a sound way to promote understandingand learning. The enormity of the material necessitatedan alternative additional method to disseminate the fig-ures and tables of the reviewed literature. Using a CD-ROM seemed a sensible method to provide this materialto you.

Your comments on the text (and the format) are wel-come. However, the true test of the success of the prod-uct is whether these guidelines are actually adopted indaily practice and the extent to which the content formsa basis of clinical practice in stroke rehabilitation.

Elliot J. Roth, MDCo-Editor-in-Chief

Topics in Stroke RehabilitationChicago, IL, USA

January 2006

Ottawa Panel Evidence-Based ClinicalPractice Guidelines for Post-Stroke

RehabilitationThe Ottawa Panel*

Top Stroke Rehabil 2006;13(2):1269 2006 Thomas Land Publishers, Inc.www.thomasland.com

1

Background and Purpose: The purpose of this project was to create guidelines for 13 types of physical rehabilitationinterventions used in the management of adult patients (>18 years of age) presenting with hemiplegia or hemiparesisfollowing a single clinically identifiable ischemic or hemorrhagic cerebrovascular accident (CVA). Method: Using CochraneCollaboration methods, the Ottawa Methods Group identified and synthesized evidence from comparative controlledtrials. The group then formed an expert panel, which developed a set of criteria for grading the strength of the evidenceand the recommendation. Patient-important outcomes were determined through consensus, provided that theseoutcomes were assessed with a validated and reliable scale. Results: The Ottawa Panel developed 147 positiverecommendations of clinical benefit concerning the use of different types of physical rehabilitation interventions involved inpost-stroke rehabilitation. Discussion and Conclusion: The Ottawa Panel recommends the use of therapeutic exercise,task-oriented training, biofeedback, gait training, balance training, constraint-induced movement therapy, treatment ofshoulder subluxation, electrical stimulation, transcutaneous electrical nerve stimulation, therapeutic ultrasound,acupuncture, and intensity and organization of rehabilitation in the management of post stroke. Key words: clinical practiceguidelines, CVA, epidemiology, evidence-based practice, outcomes, physical rehabilitation, stroke

Stroke is the third cause of mortality in NorthAmerica.1 Although approximately twothirds of stroke patients survive an initialstroke, nearly one half of survivors have physicaldisabilities as a result.2 Furthermore, while severestroke incidence has decreased, milder strokeincidence with minimal and moderate deficits hasincreased. The individuals surviving a stroke

require rehabilitation that includes varying degreesof medical care, rehabilitation, nursing, and otherhealth professional care.3 Stroke survivors presentsensorimotor, musculoskeletal, perceptual, andcognitive system deficits.4 Their impairments,disabilities, and handicaps can lead to devastatingpersonal consequences as well as consequences forthe health care system and society at large.

Special Issue

*Ottawa EBCPGs Development Group: Lucie Brosseau, PhD,1

George A. Wells, PhD,2,3 Hillel M. Finestone, MD,6 Mary Egan,PhD,1 Claire-Jehanne Dubouloz, PhD,1 Ian Graham, PhD,4 LynnCasimiro, MA, Vivian A. Robinson, MSc,3 Martin Bilodeau,PhD,1 and Jessie McGowan, MLIS.3

External Panel Members: Robert Teasell, MD,10 JohanneDesrosiers, PhD,5 Susan Barreca, MSc,8 Lucie Laferrire, MHA,9

Joyce Fung, PhD,7 Hlne Corriveau, PhD, MHA,5 GordonGubitz, MD,11 Michael Sharma, MD,9 and Mr. S. U.12

Assistant Manuscript Writers: Amole Khadilkar, MD,1 KarinPhillips, MA,1 Nathalie Jean,1 Catherine Lamothe,1 Sarah Milne,MSc,1 and Joanna Sarnecka, MSc.1

1School of Rehabilitation Sciences, Faculty of Health Sci-ences, University of Ottawa, Ottawa, Ontario, Canada; 2De-partment of Epidemiology and Community Medicine, Univer-

sity of Ottawa, Ottawa, Ontario, Canada; 3Centre for GlobalHealth, Institute of Population Health, Ottawa, Ontario,Canada; 4School of Nursing Sciences, Faculty of Health Sci-ences, University of Ottawa, Ottawa, Ontario, Canada; 5Re-search Centre on Aging and Sherbrooke University,Sherbrooke, Qubec, Canada; 6Sisters of Charity of OttawaHealth Service, Ottawa, Ontario, Canada; 7Department ofPhysical Therapy, McGill University, Montreal, Qubec,Canada; 8Hamilton Health Sciences, Hamilton, Ontario,Canada; 9Regional Stroke Centre, Ottawa Hospital, Ottawa(Ontario), Canada; 10University of Western Ontario, London,Ontario, Canada; 11Division of Neurology, Dalhousie University,Halifax (Nova Scotia), Canada; 12Patient who had a stroke.

2 TOPICS IN STROKE REHABILITATION/SPRING 2006

Post-stroke physical rehabilitation interventionshave been used to reduce pain and spasticity, aswell as to increase range of motion (ROM), muscleforce, mobility, walking ability, functional status,physical fitness, and quality of life. Post-strokephysical rehabilitation interventions are mostlynoninvasive interventions that present very fewadverse side effects and contraindications ascompared with a large number of pharmacologicinterventions.

Despite the fact that significant progress hasbeen made in the clinical management of strokeover the last decade, there is an urgent need forphysicians, nurses, physiotherapists, occupationaltherapists, and other rehabilitation specialists toprovide the most efficient and effective treatmentsfor their patients.

Evidence-based clinical practice guidelines(EBCPGs) have been defined as systematically de-veloped statements to help practitioners and pa-tients with decisions about appropriate health carefor specific clinical circumstances.5 EBCPGs are arapidly emerging technology with considerablepotential to alter the clinical decision-making pro-cess in fundamental ways. The appropriate use ofguidelines has been demonstrated to improve boththe process of care and patient health outcomes.6

EBCPGs allow stroke patients to benefit maximallyfrom the physical rehabilitation treatment thatthey are receiving.

There are currently many systematic reviewsand meta-analyses on the effectiveness of post-stroke physical rehabilitation interventions in thescientific literature. (Summarized comparative re-sults of these reviews are included in the Discus-sion section.) Trials on the efficacy of the followingtypes of therapeutic exercises for stroke survivorshave been systematically reviewed: physical fit-ness,7,8 therapeutic exercise,914 task-orientedtraining,15 progressive strengthening exercise,16,17

robot-aided training,18,19 and constraint-inducedmovement therapy.2023 Four meta-analyses havebeen published2427 for the effect of the intensity ofrehabilitation following stroke, while reviews ondifferent aspects of gait training,28 such as the useof the treadmill combined with body support,2935

have also been done. Both Barclay-Goddard35 andPollock36 have systematically reviewed balance

training. Furthermore, the efficacy of organizationand intensity of stroke rehabilitation has been ex-amined through systematic reviews.3744

Several meta-analyses, systematic reviews, andliterature reviews have been conducted over thelast 7 years on the effectiveness of EMG biofeed-back (EMG-BFB). One examined EMG-BFB forneuromuscular reeducation,45 and others lookedat EMG-BFB for the improvement of upper ex-tremity function.4648 Finally, Moreland et al.49 ex-amined the use of EMG-BFB to improve lowerextremity function after stroke. These publica-tions, though recent, require updating because ofthe rapidly growing number of scientific articlespublished on the effectiveness of EMG-BFB.50,51

Moreover, one of the studies45 has been criticizedfor failing to perform a sensitivity analysis on thecontrol group.49 Follow-up of drop-outs was alsolacking.49 EMG-BFB constitutes a small but effec-tive part of lower extremity physical rehabilitationin stroke patients.

Functional electrical stimulation (FES) has alsobeen systematically reviewed for post-stroke pa-tients.5260 A recent review was conducted on themanagement of shoulder pain and subluxation.6163 Meta-analyses on post-stroke pain managementwere done for acupuncture as an adjuvant therapyin stroke rehabilitation.6467 However, several trialsof electroanalgesia could have been added to thisreview.68,69 To our knowledge, no published sys-tematic review exists concerning the efficacy oftherapeutic electrotherapy modalities.

It is evident that the post-stroke physical reha-bilitation literature has been exhaustively re-viewed. However, the methodology used in thesereviews needs to be standardized (e.g., selectioncriteria) and quantified (e.g., Cochrane Collabora-tion), and the results of most of these reviews alsoneed to be updated in order to be included in thedevelopment of EBCPGs.

Several multidisciplinary EBCPGs have beenpublished on post-stroke rehabilitation.7078 How-ever, the Agency for Health Care Policy andResearchs (AHCPRs) EBCPGs were developed forlimited clinical practice areas.70 They did not pro-vide a clear definition of physiotherapy or reviewspecific physical rehabilitation interventions. Theyalso failed to use a rigorous grading system to

Clinical Practice Guidelines 3

assess the evidence. Guidelines were based mainlyon committee opinions and have not been recentlyupdated.79 Other EBCPGs7278 are also available forrehabilitation specialists. Although the review ofthe literature and selection of stroke topics is ex-haustive in these guidelines, the evaluation of theevidence is based upon descriptive conclusions ofthe primary studies rather than a quantitativeanalysis of the raw data. These guidelines use agrading system that takes the research design ofthe studies into account, but not the clinical sig-nificance of the outcomes. Except for the guide-lines from the Heart and Stroke Foundation ofOntario,71 these EBCPGs do not base their assess-ment of the level of evidence on a quantitativesynthesis using the raw data of the studies of inter-est, such as proposed by the Cochrane Collabora-tion methodology. These analyses are also notpooled to specific outcome measures. The conclu-sions of most of these EBCPGs concerning theeffectiveness of the selected post-stroke interven-tions are often imprecise and difficult to apply tothe daily practice of rehabilitation practitioners.The advantage of the proposed Ottawa PanelEBCPGs8082 on post-stroke physical rehabilitationinterventions is that they offer graded, quantita-tive,83 and high-quality84 recommendations thatindicate the treatment time for which a specificintervention is optimally effective for a specificoutcome for a particular stroke population.

The generally positive (small-to-large effect sizesfrom quantitative reviews) results from the recentmeta-analyses coupled with the lack of up-to-date,rigorously developed EBCPGs on physical post-stroke rehabilitation interventions suggest theneed for the development of better quality EBCPGsfor these interventions. Furthermore, evidencesuggests that quality of care can be improvedthrough the use of EBCPGs.8589 The purpose ofdeveloping these guidelines is to promote the ap-propriate use of various physical rehabilitation in-terventions in the management of stroke survivors.These guidelines are aimed at various users, in-cluding physical therapists, occupational thera-pists, physicians, and patients. This article dis-cusses only post-stroke physical interventionssuch as therapeutic exercises, task-oriented train-ing, biofeedback, gait training, balance training,

sensory interventions, constraint-induced move-ment therapy (CIMT), treatment of shoulder sub-luxation, electrical stimulation, transcutaneouselectrical nerve stimulation (TENS), therapeuticultrasound, acupuncture, and intensity and orga-nization of rehabilitation.

Method

The development process of these EBCPGs wassimilar to that of the Philadelphia Panel90 and toprevious Ottawa Panel publications,81,82 exceptthat a different target population was used. Briefly,the Ottawa Methods Group (OMG), a group of 10methodologists with experience in developingEBCPGs, asked professional associations inter-ested in the care of stroke patients to suggest indi-viduals with both clinical expertise in the manage-ment of stroke and familiarity with EBCPGs. Fromamong the suggestions, the OMG chose nine ex-perts (R.T., G.G., J.D., J.F., H.C., S.B., L.L., M.S.,S.U.) to serve as panel members. The professionalexperts were recruited from multidisciplinary dis-ciplines such as physical medicine, neurology, oc-cupational therapy, and physical therapy. Severalexperts (R.T., G.G., J.D., J.F., S.B., L.B., A.H.) aremembers of the Canadian Stroke Network,91 whilesome had already developed post-stroke rehabili-tation EBCPGs (R.T., S.B.). The Ottawa Panel con-sisted of these nine experts, in addition to all themembers of the OMG.

The OMG assembled a research and supportstaff with expertise in meta-analyses, stroke reha-bilitation interventions, research methods, or thedevelopment and assessment of EBCPGs. TheOMG then established an a priori set of inclusioncriteria for the study designs, subject samples, in-terventions, and outcomes to allow the researchstaff to select the most relevant material as evi-dence for the effectiveness of various rehabilitationinterventions for post-stroke patients. The OMGalso reviewed the inclusion criteria to ensure thatthe approach to the study selection was reproduc-ible and systematic. This a priori protocol guidedseparate systematic reviews of the literature foreach intervention. The OMG also made sure thatthe Ottawa Panel EBCPGs were methodologically

4 TOPICS IN STROKE REHABILITATION/SPRING 2006

developed at high level of quality, according toAGREE (www.agreecollaboration.org) criteria.84

The research staff reviewed articles and createddraft evidence tables, which the nine clinical expertsreceived in preparation for their consensus meetingwith the OMG. These tables were used as the basisfor making the Ottawa Panel recommendations.

Target population

The target population was adult patients (>18years of age) presenting with hemiplegia or hemi-paresis following a single clinically identifiable is-chemic or hemorrhagic CVA. The patients had tobe medically stable and able to follow simple in-structions and to interpret and respond to feed-back signals. The mean duration since stroke onsetvaried from hyper-acute (the first 12 hours), acute(first week following a stroke), subacute (from thefirst to 6th week), and post-acute (from 6 weeks to6 months) to chronic (from 6 months) as definedby the Canadian Stroke Network (Appendix 3:Characteristics of Included Studies).91

Patients who had been identified as having mul-tiple CVAs, other neurological problems (e.g.,Parkinsons, brain tumors, traumatic brain injury),subarachnoid hemorrhages, or subdural hemato-mas were excluded because of the numerous andvaried associated signs and symptoms. Studies thatincluded patients with bilateral neurological signswere also excluded. Further exclusion criteria in-cluded studies whose patients presented with oneof the following conditions: (1) cancer or otheroncological conditions, (2) cardiac conditions, (3)dermatologic conditions, (4) healthy normal sub-jects, (5) serious cognitive deficits or severe com-munication problems, (6) major medical problemsthat could interfere with the rehabilitation pfOcessor incapacitate functional status, or (7) psychiatricconditions. Further inclusion and exclusion crite-ria are exhibited in Table 1.

Literature search

The library scientist developed a structured lit-erature search based on the sensitive search strat-egy recommended by The Cochrane Collabora-tion92 and modifications to that strategy proposed

by Haynes et al.93 The Cochrane Collaborationmethod minimizes bias through a systematic Z -proach to the literature search, study selection,and data extraction and synthesis. The search wasorganized around the condition and interventionsrather than the outcomes because it was an a priorisearch. Thus, we had no control over the outcomesthat the authors of the primary studies decided tomeasure (Appendix 1: Literature Search Results).

The library scientist expanded the search strat-egy to identify case control, cohort, and non-ran-domized studies and conducted the search in theelectronic databases of MEDLINE, EMBASE, Cur-rent Contents, the Cumulative Index to Nursingand Allied Health (CINAHL), and the CochraneControlled Trials Register up to December 2004.She also searched the registries of the CochraneField of Rehabilitation and Related Therapies, theCochrane Musculoskeletal Group, the Physio-therapy Evidence Database (PEDro), and the Uni-versity of Ottawa EBCPGs Web site. Finally, shesearched the reference lists of all of the includedtrials for relevant studies and contacted contentexperts for additional studies.

In the first round of study inclusion or exclu-sion, two trained independent reviewers appraisedthe titles and abstracts of the literature search,using a checklist with the a priori defined selectioncriteria (Table 1). For each pair of reviewers, indi-viduals independently read the title and abstract ofeach article and created a list of all of the articles inthe database along with a reason for either includ-ing or excluding each article. If the reviewers wereuncertain about a particular article after havingread the abstract, they ordered the article and readit in full before making a determination. Beforedeciding whether to include or exclude the article,a comparison of their individual lists was per-formed. A senior reviewer, a methodologist and aclinical expert (L.B.), checked the two indepen-dent lists of articles and the reasons for inclusionor exclusion to determine potential inconsisten-cies. Seven percent of the abstracts needed theconsultation of the senior reviewer and an addi-tional review of the problematic article. For thesecond round of the inclusion and exclusion pro-cess, the pairs of reviewers retrieved articles se-lected for inclusion from the first round and inde-

Clinical Practice Guidelines 5

pendently assessed the full articles for inclusion orexclusion in the study. Using predetermined ex-traction forms, the pairs of reviewers indepen-dently extracted from included articles data on thepopulation characteristics, details of the interven-tions, trial design, allocation concealment, andoutcomes. The pairs of reviewers assessed themethodological quality of the studies using theJadad Scale,83,94 a 5-point scale with reported reli-ability and validity that assigns 2 points each forrandomization and double blinding and 1 pointfor description of withdrawals. The reviewers re-solved differences in data extraction and qualityassessment through consensus with the senior re-viewer. This consensus served to support the reli-ability of data obtained with the article selectionprocess.

Study inclusion/exclusion criteria

The inclusion/exclusion criteria were basedupon previous criteria used by the PhiladelphiaPanel.90 This list of criteria, which had been cre-ated for multiple diagnoses, was adapted and ap-proved by the OMG for use with patients poststroke (Table 1).

All original comparative controlled studies thatevaluated relevant physical rehabilitation interven-tions in stroke patients were included: randomizedcontrolled trials (RCTs), controlled clinical trials(CCTs),* cohort studies, and case-control studies.Crossover studies were included, but to avoid po-tential confounding carry-over effects the datafrom only the first part of the study (before cross-ing) was analyzed. Studies where patients servedas their own controls were excluded. No limita-tions based on methodological quality were im-posed a priori with regard to the selection of com-parative controlled studies; however, the quality ofthe studies was considered when grading the rec-ommendations resulting from our analysis.

Uncontrolled cohort studies (studies with nocomparison group) and case series were excluded,as were eligible studies with greater than a 20%

drop-out rate or a sample size of fewer than 5patients per group. Trials published in languagesother than French and English were not analyzed,because of the additional time and resources re-quired for translation. Abstracts were excluded ifthey contained insufficient data for analysis andadditional information could not be obtained fromthe authors. For further exclusion criteria, seeTable 1.

Post-stroke rehabilitation interventions

Post-stroke rehabilitation interventions wereidentified as therapeutic exercises, task-orientedtraining, biofeedback, gait training, balance train-ing, sensory information, CIMT, treatment ofshoulder subluxation, electrical stimulation,TENS, therapeutic ultrasound, and acupuncture.Intensity and organization of rehabilitation wasalso included as an intervention related to strokerehabilitation.

Post-stroke rehabilitation interventions relatedto therapeutic exercises were identified as aerobictraining, resistance training, passive range of mo-tion exercises, proprioceptive neuromuscular fa-cilitation, Bobath technique, kinetron, and the useof the overhead pulley.95 They are defined as fol-lows. Aerobic training is considered to be activitiesto increase endurance and cardiovascular func-tion. Resistance training was defined as active exer-cise done against a resistance. Passive range of mo-tion exercises were defined as physiologicalmobilization done by the therapist without anyeffort from the patient. Proprioceptive neuromuscu-lar facilitation was identified as the use of mostlyreflex-inhibiting patterns. Bobath technique was de-fined as a neurodevelopmental technique usinginhibitory posture and movement to inhibit spas-ticity and synergies, while facilitating normalmovements. Kinetron training was defined as train-ing with this resistive lower extremity machine,usually in isokinetic mode.

Rehabilitation interventions related to task-ori-ented training were identified as treatments thatinvolved dividing activities of daily living intocomponent parts. Individual components of thelarger task were then practiced until the patientwas able to complete the component adequately.

*Controlled clinical trials are considered the same as randomizedcontrol trials (RCTs). However, according to the Jadad Scale,94 CCTs areeither not randomized or have not been appropriately randomized.

6 TOPICS IN STROKE REHABILITATION/SPRING 2006

Component parts were then combined, and theoverall skill was practiced with repetition. Anyintervention that divided required tasks into indi-vidual skills was included.95 We included taskssuch as seated reaching tasks, adapted games, andrepetitive elbow joint movements.

Post-stroke rehabilitation interventions relatedto biofeedback were identified as EMG-BFB, EMG-biofeedback-relaxation training, rhythmic posi-tional biofeedback, audio and visual feedback,video feedback, and force feedback. EMG-BFB wasdefined as an intervention that allows a patient tomonitor his or her muscle activity through elec-trodes with a visual or audible feedback signal.95

Rhythmic positional biofeedback was defined as ausual biofeedback intervention with auditory orvisual stimuli aimed at increasing the rhythm ofmovement. Audio and visual feedback were identi-fied as any cue received by the patient during orafter the exercise. Video feedback was considered tobe a visual cue from a monitor after the action wasdone. Force feedback was defined as feedback re-lated to the moment of force.

Rehabilitation interventions associated with gaittraining were identified as treadmill training,overground training, body weight support train-ing, brace-assisted walking, electrogoniometricfeedback training, FES, rhythmic auditory facilita-tion training, and functional lower extremity train-ing. Treadmill training was defined as ambulationon a treadmill adjusted to patients comfortablewalking speed or highest speed as possible for thepatient. Overground training was defined as gaittraining on an even surface with propulsion for-ward, backward, and sideways or going up anddownstairs. Body weight support training was de-fined as treadmill training, while an overhead har-ness supported a percentage of the body weight.Brace-assisted walking was defined as use ofhemibar and ankle-foot orthosis (AFO) or anyother type of brace if necessary. Electrogoniometricfeedback training was defined as auditory feedbackduring gait training when patient was compensat-ing in hyperextension or flexion. FES was definedas electrical stimulation of a specific muscle ornerve such as tibialis anterior or peroneal nerve fora functional purpose to improve swing phase orstance phase during gait. Rhythmic auditory facilita-tion training was defined as imposed rhythm to

improve gait rhythm and frequency. Functionallower extremity training was defined as functionaltasks such as sitting, standing, climbing stairs,transfers, and gait with a focus on the recovery ofstability and gait performance.

Rehabilitation interventions related to balancetraining were identified as any intervention thatcontributes to the enhancement of equilibriumand balance in post-stroke patients. We includedinterventions such as base of support training andplatform training.

Rehabilitation interventions related to sensoryinterventions were identified as any retraining ofthe sensory and visuo-spatial function to correctposture and perceptual problems after stroke,95

such as passive vestibular stimulation, perceptuallearning exercises, and rocking chair stimulation.

Rehabilitation interventions associated withCIMT were defined as the restriction of thenonparetic upper extremity by a sling or handsplint to encourage the use of the paretic limb.Functional exercises were given to the patient toimprove the function of the affected arm.

Post-stroke rehabilitation interventions relatedto treatment of shoulder subluxation were identifiedas FES, supports methods, strapping, and shoul-der positioning. FES was defined as electricalstimulation of a specific muscle or nerve such assupraspinatus or middle deltoid with functionalpurpose resulting in the reduction of shouldersubluxation. Support methods were defined as anyuse of an external support such as orthosis or slingto prevent shoulder subluxation. Strapping methodswere defined as strapping used to keep the gleno-humeral joint in normal position. Shoulder position-ing was defined as a position induced by the phys-iotherapist to protect the structures around theweak hemiplegic shoulder in order to avoid shoul-der pain and shoulder subluxation.

Post-stroke rehabilitation interventions relatedto electrical stimulation were identified as FES,neuromuscular electrical stimulation (NMES),positional feedback stimulation training, EMG-triggered electrical muscle stimulation, andTENS. FES was defined as the electrical stimula-tion of a specific muscle or nerve such as tibialisanterior and gastrocnemius for a functional pur-pose, such as gait training. Neuromuscular electri-cal stimulation (NMES) was defined as electrical

Clinical Practice Guidelines 7

stimulation of a specific muscle or nerve such asthumb flexors and extensors to help trigger ner-vous fibers and achieve motor recovery. Positionalfeedback stimulation training was defined as audi-tory and visual feedback during training toachieve a target position of the joint. EMG-trig-gered electrical muscle stimulation was defined aselectrical stimulation of a muscle triggered byEMG activity of this muscle.

TENS was defined as a form of electrical stimula-tion that triggers nervous endings to inhibit themessage of pain. TENS is identified as being givenat high- and low-intensity levels.

Ultrasound was identified as an electrophysicalmodality using an ultrasonic wave to treat a spe-cific area, usually for pain, and indirectly for ROM.

Post-stroke rehabilitation interventions relatedto acupuncture were identified as any treatmentusing needles to stimulate specific anatomicalpoint with the hands or with electrical stimulation.

Post-stroke rehabilitation interventions relatedto the intensity and organization of rehabilitationwere identified as examining the rate, frequencyand rigor of any physical rehabilitation interven-tion or combination of interventions in the treat-ment of post-stroke patients. 95 Interventions suchas stroke unit care, enhanced physical therapy,enhanced occupational therapy, enhanced upperextremity treatment, and intensive outpatientphysiotherapy rehabilitation were included as partof intensity and organization of post-stroke reha-bilitation.

Acceptable comparisons were placebo, no treat-ment, or use of educational pamphlets. Studiesdesigned with a comparison of two interventionsinstead of treatment versus control were includedas long as both interventions respected the inclu-sion criteria. Concurrent therapies (such as medi-cation) were accepted only if they were providedto both the experimental and control groups.Study selection was not restricted by the cost,complexity, or general availability of equipmentand resources required to carry out the interven-tions under investigation.

Outcomes

The outcomes were selected and based upon theWorld Health Organizations96 (WHOs) new pro-

posal of the International Classification of Function-ing, Disability and Health, which involved the con-cepts of body function, body structure, activitiesand participation, and environmental factors. Thea priori outcomes were classified according to twoWHO concepts:

(1) Body function: pain reduction, musclestrength, motor function/motor recovery, ROM,postural status, balance status, gait status, cadence,stride length, sensory status, spasticity/muscletone, global physician assessment, global patientassessment, and cardiopulmonary function.

(2) Activities and participation: walkingspeed, walking distance, endurance, functionalstatus, patient adherence, patient satisfaction,length of stay, discharge disposition, quality of life,and return to work.

Studies were included if any one of the afore-mentioned outcomes was measured. A positiverecommendation was made only if a specific inter-vention was effective for an outcome, as measuredby a validated scale. Psychological outcomes suchas depression were excluded (Table 1).

Statistical analysis

The data were analyzed using Review ManagerSoftware.97 Continuous data, data with a poten-tially infinite number of possible values along acontinuum, 98 were analyzed using the weightedmean differences (WMDs) between the interven-tion and control groups at the end of the study,where the weight is the inverse of the variance. AWMD is a method of meta-analysis used to com-bine measures on continuous scales (such asweight), where the mean, standard deviation andsample size in each group are known. 98 Dichoto-mous data or data with only two classifications98

were analyzed using relative risks. According toCochrane, the relative risk is the ratio of risk inthe intervention group to the risk in the controlgroup. The risk (proportion, probability, or rate) isthe ratio of people with an event in a group to thetotal in the group. 98

Heterogeneity (i.e., variability or difference inestimated effects between studies) was tested usingthe chi-square statistic. We tested data heterogene-ity across the results of different included studies.When heterogeneity was not significant, fixed-ef-

8 TOPICS IN STROKE REHABILITATION/SPRING 2006

fect models were used. A fixed-effect model is a statisti-cal model that stipulates that the units under analysis(e.g., participants in a meta-analysis study) are theones of interest and thus constitute the entire popula-tion of units. Fixed-effect models were used to gener-alize data across the included studies. Random-ef-fects models include both within-study samplingerror (variance) and between-study variation in theassessment of the uncertainty (confidence interval) ofmeta-analysis results. Such random-effects modelswere used when heterogeneity was significant. Allfigures were created using Cochrane Collaborationmethodology (www.cochrane.org). The square inTE-Figure 1A. illustrates the WMD between the twogroups, when they are compared for a specific out-come of interest. The horizontal line represents thestandard deviation of the WMD. If the standard de-viation line touches the central vertical line of thegraph, the confidence interval contains a zero and thedifference between the two groups is not statisticallysignificant. For example, in TE-Figure 1A that illus-trates the comparison between aerobic training and acontrol group where gait speed at end of treatment isthe outcome measure, the gait speed of the groupreceiving the aerobic training is not statistically differ-ent from that of the control. Based on previous Phila-delphia and Ottawa Panels consensus,8082 clinicalimprovement for all interventions studied by theOttawa Panel was defined as 15% improvement,relative to a control.

To determine clinical improvement, the abso-lute benefit and relative difference in the changefrom baseline were calculated. Absolute benefit wascalculated as the improvement in the treatmentgroup minus the improvement in the controlgroup, maintaining the original units of measure-ment. The relative difference (RD) was calculated asthe absolute benefit divided by the baseline mean(weighted for the intervention and controlgroups). For dichotomous data, the relative per-centage of improvement was calculated as the differ-ence in the percentage of improvement betweenthe intervention and control groups.80

However, during this meta-analysis, four specialcases of calculation appeared where new formulaswere needed to calculate the RD in the changefrom baseline. The first case occurred when thebaseline values were not available for an outcome.The second scenario involved an outcome that was

measured as a change from baseline, where thescale of measurement was known but baselinevalues were absent. The third case encounteredwas also where the outcome was measured as achange from baseline but both the baseline valuesand the measurement scales were either not avail-able or were nonexistent (e.g., no measurementscales exist for strength). Finally, in the fourthcase, where the baseline mean was given as 0, asum of 1 was added to all the values in the formulaof clinical relevance, based upon the followingassumption: the mean of scale + 1 = mean oforiginal scale + 1 and SD of scale + 1= SD oforiginal scale. The new formulas used to calculatethe relative difference in change from baseline foreach of the four aforementioned scenarios aregiven in Appendix 2.

The recommendations were graded by their level(I for RCTs, II for nonrandomized studies) andstrength (A, B, C+, C, or D) of evidence. Evidencefrom one or more RCTs of a statistically significant,clinically important benefit (>15%) was necessaryfor a grade A recommendation. A grade B recom-mendation was given for a statistically significant,clinically important benefit (>15%), if the evidencewas from observational studies or CCTs. Evidenceof clinical importance (>15%) but not statisticalsignificance earned a grade C+ recommendation. Agrade C recommendation was given to those inter-ventions where an appropriate outcome was mea-sured in a study that met the inclusion criteria, butno clinically important difference and no statisticalsignificance were shown. Evidence from one ormore RCTs of a statistically significant benefit fa-voring the control group (

Clinical Practice Guidelines 9

the scientific literature by an existing validationstudy but that provided useful information in thestudies were insufficient to warrant a grade A or Brecommendation.

Reviewing the guidelines

The guidelines were sent to the external expertsfor review. To judge the clinical usefulness of theguidelines, the positive recommendations were alsosent to practitioners for feedback. Practitioners wereasked four questions for each guideline: whether therecommendation was clear, whether the practitio-ners agreed with the recommendation, whetherthey felt that the literature search on the differentintervention of rehabilitation was relevant and com-plete, and whether the results of the trials in theguidelines were interpreted according to the practi-tioners understanding of the data. Their questionsand comments were carefully addressed to improvethe clarity of the final guidelines.

Results of literature search

The initial literature search in 2002 identified1,533 potential articles on post-stroke rehabilita-tion interventions. From these, 148 articles ontherapeutic exercises for post-stroke rehabilitationwere initially identified. After many updated litera-ture searches that ended in December 2004, 151articles were considered potentially relevant and,of these, 71 articles met the selection criteria andwere included. From the selected articles, we cre-ated the following five subcategories of therapeuticexercise: therapeutic exercises, task-oriented train-ing, CIMT, balance training, and sensory interven-tion. In the end, 56 trials were considered fortherapeutic exercises. From these, 29 trials met theinclusion criteria and were then included (Appen-dix 3A) and 27 trials were excluded for differentreasons (TE-Table 3). In task-oriented training, 39articles were considered as potentially relevant.From these, 17 were included (Appendix 3B) and22 were excluded (TOT-Table 37). Initially, 131articles were identified for biofeedback. After thelast search in December 2004, 66 were consideredpotentially relevant. From these, 19 were included(Appendix 3C) and 45 were excluded for differentreasons (BFB-Table 57). Also, 433 potential ar-

ticles on gait training for post-stroke rehabilitationwere initially identified. By the end of December2004, 50 of these were considered potentially rel-evant based upon the selection criteria. Nineteenof these articles relating to gait training met theselection criteria and were included (Appendix3D). The other 23 trials were excluded from thefinal selection for various reasons (Gait-Table 73).

For balance training, 16 articles were consideredafter the last search. From these, 11 met the inclu-sion criteria and were kept (Appendix 3E) and 4were excluded (BT-Table 96). From the 23 poten-tially relevant articles concerning sensory interven-tions at the end of 2004, 9 were included (Appen-dix 3F) and 14 were excluded (SI-Table 107).CIMT included 5 articles (Appendix 3G), while13 trials were excluded (CIMT-Table 117) fromthe various updated searches leading up to Decem-ber 2004 (N = 17 potential articles). Twenty-twotrials were considered relevant in the treatment ofshoulder subluxation. Eight were included (Appen-dix 3H) and the other 13 were excluded for vari-ous reasons (SH-Table 123).

The initial search found 368 studies for electro-therapy. From the last search of 2004, 75 trials wereconsidered potentially relevant. Thirty studies wereincluded and 45 were excluded for different rea-sons. Electrotherapy was then separated into morespecific interventions: electrical stimulation, TENS,therapeutic ultrasound, and, finally, acupuncture.In electrical stimulation, 41 articles were consideredpotentially relevant. Eleven of them ended up beingincluded (Appendix 3I) and 26 were excluded (ES-Table 135). TENS had 11 potentially relevant trials.Six finally met the inclusion criteria and were in-cluded (Appendix 3J) and six were excluded(TENS-Table 148). Four articles were consideredpotentially relevant for therapeutic ultrasound. How-ever, only one of these was included (Appendix3K) and the remaining three were excluded (US-Table 157). Acupuncture was found to have 19potential articles. Seven were included (Appendix3L) and 11 were excluded (AC-Table 160). Inten-sity and organization of rehabilitation initially had 272potential articles. By the end of 2004, 156 wereconsidered as potentially relevant. Fifty-six werefinally included with respect to the inclusion criteria(Appendix 3M) and 102 were excluded for differ-ent reasons (IR-Table 167).

10 TOPICS IN STROKE REHABILITATION/SPRING 2006

Results

Clinical practice guidelines for therapeutic exercises

Aerobic training versus control, level I (RCT):Grade A for cardiopulmonary function (expira-tion per minute [VE]), muscle power, and func-tional status (walking) at end of treatment, 10weeks (clinically important benefit demonstrated);grade C+ for gait speed at end of treatment, 10weeks (clinically important benefit demonstratedwithout statistical significance); grade C for car-diopulmonary function (maximal heart rate, VO

2

max, VCO2

max) and motor function at end oftreatment, 10 weeks, and for functional status(Frenchay Activities Index [FAI]: global, socialoutings, walking outside) at end of treatment, 6months (no benefit demonstrated); grade D forfunctional status (FAI: light housework activities)at end of treatment, 6 months (no benefit demon-strated but favoring control). Patients with sub-acute and chronic stroke.

Aerobic individualized program training ver-sus control, level I (RCT): Grade A for physicalfitness (highest test stage completed of the stresstest and maximal workload) and mobility (stairclimbing) at end of treatment, 8 weeks (clinicallyimportant benefit demonstrated); grade C+ formobility (walking distance) at end of treatment, 8weeks (clinically important benefit demonstratedwithout statistical significance); grade C for car-diovascular function (maximal heart rate, decreaseof resting heart rate and decrease of resting systolicand diastolic blood pressure), gait speed, andfunctional status at end of treatment, 8 weeks (nobenefit demonstrated). Patients with subacutestroke.

Proprioceptive neuromuscular facilitation(PNF) for upper extremity versus standard cus-tomary muscle training, level I (RCT): Grade Cfor functional status and upper extremity musclestrength at end of treatment, 6 weeks (no benefitdemonstrated); grade D for mobility at end oftreatment, 2, 4, and 6 weeks, ROM of the wrist atend of treatment, 6 weeks (no benefit demon-strated but favoring control); grade D+ for ROM ofthe ankle at end of treatment, 6 weeks (clinicallyimportant benefit favoring control demonstrated

without statistical significance). Patients with post-acute stroke.

PNF versus Bobath approach training, level I(RCT): Grade C for mobility at end of treatment,2, 4, and 6 weeks, ROM of the wrist and ankle atend of treatment, 6 weeks (no benefit demon-strated). Patients with subacute stroke.

Bobath approach versus standard customarymuscle training, level I (RCT): Grade C+ motorfunction (Sodring Motor Evaluation Scale[SMES]: upper extremity) at follow-up, 4 years,quality of life (Nottingham Health Profile [NHP]global) at follow-up, 1 year and 4 years, andquality of life (NHPloss of energy) at end oftreatment, 3 months (clinically important benefitdemonstrated without statistical significance);grade C for mobility at end of treatment, 2, 4,and 6 weeks, motor function (SMES: lower ex-tremity and trunk, balance, and gait, and MotorAssessment Scale) at end of treatment, 3 months,and follow-up, 1 year and 4 years, motor function(SMES: upper extremity) at end of treatment, 3months, and follow-up, 1 year, functional status(Barthel Index) at follow-up, 4 years (no benefitdemonstrated); grade D for ROM of the wrist andankle at end of treatment, 6 weeks, functionalstatus (Barthel Index) at end of treatment, 3months, and follow-up, 1 year, (no benefit dem-onstrated but favoring control); grade D+ forpain relief (NHPpain) at end of treatment, 3months (clinically important benefit favoringcontrol demonstrated without statistical signifi-cance). Patients with acute and subacute stroke.

Bobath approach training versus control,level I (RCT): Grade C+ for balance standing atfollow-up, 2 and 12 weeks (clinically importantbenefit demonstrated without statistical signifi-cance); grade C for balance sitting at end of treat-ment, 4 weeks (no benefit demonstrated); grade Dfor balance sitting at follow-up, 2 weeks, and bal-ance standing at end of treatment, 4 weeks (nobenefit demonstrated but favoring control); gradeD+ for balance sitting at follow-up, 12 weeks(clinically important benefit favoring control dem-

Clinical Practice Guidelines 11

onstrated without statistical significance). Patientswith subacute stroke.

Progressive resistance training versus activetraining for the lower extremity, level I (RCT):Grade A for functional status at end of treatment,1 month (clinically important benefit demon-strated). Patients with post-acute stroke.

Active training for the lower extremity versuscontrol (no exercise), level I (RCT): Grade D forfunctional status at end of treatment, 1 month (nobenefit demonstrated but favoring control). Pa-tients with post-acute stroke.

Progressive resistance training for the lowerextremity versus control, level I (RCT): GradeC+ for functional status at end of treatment, 1month (clinically important benefit demonstratedwithout statistical significance). Patients with post-acute stroke.

Progressive resistance versus no resistancetraining, level I (RCT): Grade C for motor recov-ery at end of treatment, 4 weeks and 8 weeks, andfollow-up, 6 months (no benefit demonstrated);grade D for gait endurance at end of treatment, 4weeks and 8 weeks, and follow-up, 6 months (nobenefit demonstrated but favoring control). Pa-tients with subacute stroke.

Functional task training for upper extremityversus strength training, level I (RCT): Grade C+favoring functional task training for functional sta-tus (FIM* self-care and FIM mobility), isometrictorque, grip strength, and palmar pinch at follow-up, 6.5 to 8 months, and lateral pinch at end oftreatment, 4 weeks, and follow-up, 6.5 to 8months, and grade C+ favoring strength trainingfor grip strength and palmar pinch at end of treat-ment, 4 weeks (clinically important benefit dem-onstrated without statistical significance); grade Cfunctional status (FIM self-care and FIM mobility)at end of treatment, 4 weeks, ROM upper extrem-ity, pain relief, sensory function upper extremity,motor function upper extremity, functional status

(Functional Test of the Hemiparetic Upper Ex-tremity [FTHUE]) at end of treatment, 4 weeks,and follow-up, 6.5 to 8 months, isometric torqueat end of treatment, 4 weeks (no benefit demon-strated). Patients with subacute stroke.

Strength training versus control, level I (RCT):Grade A for upper extremity isometric torque atend of treatment, 4 weeks (clinically importantbenefit demonstrated); grade C+ for motor func-tion upper extremity and functional status(FTHUE) at end of treatment, 4 weeks, palmarpinch at end of treatment, 4 weeks, and follow-up,6.5 to 8 months, grip strength and lateral pinch atfollow-up, 6.5 to 8 months (clinically importantbenefit demonstrated without statistical signifi-cance); grade C for sensory function upper ex-tremity at end of treatment, 4 weeks, and follow-up, 6.5 to 8 months, functional status (FIMmobility), pain relief, and grip strength at end oftreatment, 4 weeks (no benefit demonstrated);grade D+ for lateral pinch at end of treatment, 4weeks, functional status (FIM self-care and FIMmobility), upper extremity isometric torque at fol-low-up, 6.5 to 8 months (clinically important ben-efit favoring control demonstrated without statisti-cal significance); grade D for ROM upperextremity at end of treatment, 4 weeks, and follow-up, 6.5 to 8 months, functional status (FIM self-care) at end of treatment, 4 weeks, pain relief,motor function upper extremity, functional status(FTHUE) at follow-up, 6.5 to 8 months (no benefitdemonstrated but favoring control). Patients withsubacute stroke.

Aerobic and strength versus aerobic training,level I (RCT): Grade A for cardiopulmonary func-tion and peak torque for shoulder flexors at end oftreatment, 16 weeks (clinically important benefitdemonstrated); grade C+ for peak torque for kneeflexors at end of treatment, 16 weeks (clinicallyimportant benefit demonstrated without statisticalsignificance); grade C for peak torque for shoulderextensors and peak torque for knee extensors atend of treatment, 16 weeks (no benefit demon-strated). Patients with chronic stroke.

*FIM is a trademark of Uniform Data System for Medical Rehabilita-tion, a division of UB Foundation Activities, Inc.

12 TOPICS IN STROKE REHABILITATION/SPRING 2006

Kinetron training for lower extremity versuscontrol (no Kinetron), level I (RCT): Grade D formobility at end of treatment, 5 weeks (no benefitdemonstrated but favoring control). Patients withpost-acute stroke.

Home-based exercise training versus control,level I (RCT): Grade A for change in gait speed, gaitendurance, torque (change in knee isometric exten-sors), endurance, and cardiopulmonary function atend of treatment, 12 weeks; grade C+ for motorfunction (change in Fugl-Meyer lower extremity),change in gait speed, gait endurance, and functionalstatus (physical function index), strength (change ingrip strength) at end of treatment, 12 weeks (clini-cally important benefit demonstrated without statis-tical significance); grade C for motor function(change in Fugl-Meyer upper extremity and lowerextremity), balance (Berg balance and change inBerg balance), functional status (Instrumental ADLand Barthel ADL Index), at end of treatment, 12weeks (no benefit demonstrated); grade D+ forbalance (functional reach) at end of treatment, 12weeks (clinically important benefit favoring controldemonstrated without statistical significance);grade D for torque (change in ankle isometricdorsiflexors) (no benefit demonstrated but favoringcontrol). Patients with post-acute stroke.

Skateboard versus overhead pulley trainingfor the shoulder, level I (RCT): Grade C+ for painrelief at end of treatment, 810 weeks (clinicallyimportant benefit demonstrated without statisticalsignificance). Patients with subacute stroke.

Overhead pulley versus control (passive ROMtraining for shoulder), level I (RCT): Grade D forpain relief at end of treatment, 810 weeks (nobenefit demonstrated but favoring control). Pa-tients with subacute stroke.

Passive ROM training for shoulder versusskateboard, level I (RCT): Grade C for pain reliefat end of treatment, 810 weeks (no benefit dem-onstrated). Patients with subacute stroke.

Resisted extension versus ballistic extensiontraining for the hand, level I (RCT): Grade C+ forROM at end of treatment, 2 weeks (clinically im-portant benefit demonstrated without statisticalsignificance); grade C motor function at end oftreatment, 2 weeks (no benefit demonstrated). Pa-tients with subacute and post-acute stroke.

Resisted extension versus resisted grasptraining for the hand, level I (RCT): Grade A formotor function (change in tapping) at end of treat-ment, 2 weeks (clinically important benefit dem-onstrated); grade C+ for ROM at end of treatment,2 weeks (clinically important benefit demon-strated without statistical significance); grade Cfor motor function (change in grasp/release) at endof treatment, 2 weeks (no benefit demonstrated).Patients with subacute and post-acute stroke.

Resisted extension training for the hand ver-sus control, level I (RCT): Grade A for motorfunction (change in tapping) and ROM at end oftreatment, 2 weeks (clinically important benefitdemonstrated); grade D for motor function(change in grasp/release) at end of treatment, 2weeks (no benefit demonstrated but favoring con-trol). Patients with subacute and post-acute stroke.

Ballistic extension versus resisted grasptraining for the hand, level I (RCT): Grade A formotor function (change in tapping) at end of treat-ment, 2 weeks (clinically important benefit dem-onstrated); grade C+ for ROM at end of treatment,2 weeks (clinically important benefit demon-strated without statistical significance); grade Cfor motor function (change in grasp/release) andROM at end of treatment, 2 weeks (no benefitdemonstrated). Patients with subacute and post-acute stroke.

Ballistic extension training for the hand ver-sus control, level I (RCT): Grade C+ for motorfunction (change in tapping) at end of treatment, 2weeks (clinically important benefit without statis-tical significance); grade C for ROM at end oftreatment, 2 weeks (no benefit demonstrated);grade D for motor function (change in grasp/re-lease) at end of treatment, 2 weeks (no benefitdemonstrated but favoring control). Patients withsubacute and post-acute stroke.

Resisted grasp training for the hand versuscontrol, level I (RCT), Grade C+ for ROM at end oftreatment, 2 weeks (clinically important benefitdemonstrated); grade C for motor function (changein tapping) at end of treatment, 2 weeks (no benefitdemonstrated); grade D for motor function (changein grasp/release) at end of treatment, 2 weeks (nobenefit demonstrated but favoring control). Patientswith subacute and post-acute stroke.

Clinical Practice Guidelines 13

Robot-aided training versus no robot-aidedtraining, level I (RCT) and level II (CCT): Grade Afor motor power for shoulder and elbow at end oftreatment, 5 weeks, change in motor power upperextremity at end of treatment, 6 weeks, motorfunction (Fugl-Meyer for shoulder, elbow, and co-ordination and Motor Status Score [MSS] forshoulder and elbow) at end of treatment, 5 weeks,motor function (MSS for wrist and hand) at end oftreatment, 5 weeks, and motor function (change inMSS for shoulder and elbow) at end of treatment, 6weeks, and follow-up, 3 years, motor function(MSS for wrist and hand) at end of treatment, 5weeks (clinically important benefit demonstrated);grade B for motor function (MSS for upper ex-tremity) at end of treatment, 6 weeks (clinicallyimportant benefit demonstrated); grade C+ forchange in motor power for shoulder and elbow atfollow-up, 3 years, motor function (Fugl-Meyerscale for upper extremity), motor power for upperextremity at end of treatment, 6 weeks (clinicallyimportant benefit demonstrated without statisticalsignificance); grade C for motor function (changein Fugl-Meyer for shoulder, elbow, and coordina-tion) at end of treatment, 6 weeks, functional sta-tus (FIM for upper extremity) at end of treatment,5 weeks and 6 weeks, motor function (change inFugl-Meyer for wrist and hand and change in MSSfor wrist and hand) at end of treatment, 6 weeks,and follow-up, 3 years (no benefit demonstrated);grade D for motor function (change in Fugl-Meyerfor shoulder, elbow, and coordination) at follow-up, 3 years (no benefit demonstrated but favoringcontrol). Patients with subacute-chronic stroke.

Robot-assisted versus neurodevelopmental(NDT) training, level I (RCT): Grade A forstrength (change in elbow extensors, shoulder in-ternal rotators, abductors, adductors, and flexorsstrength [%]) and functional reach (change in for-ward medial, forward, forward lateral, and lateralreach extent) at end of treatment, 2 months; gradeC+ for strength (change in shoulder external rota-tors and extensors strength [%]) at end of treat-ment, 2 months (clinically important benefit dem-onstrated); grade C for functional status (changein Barthel Index and change in FIM) at follow-up,6 months, motor function (change in Fugl-Meyershoulder and elbow) at end of treatment, 1 month,

2 months, and follow-up, 6 months, motor func-tion (change in Fugl-Meyer hand and wrist) at endof treatment, 1 month, 2 months, and follow-up, 6months, (no benefit demonstrated). Patients withchronic stroke.

Robot-aided progressive resistance trainingversus robot-aided active-assisted training, levelI (RCT): Grade C for decrease of spasticity, motorfunction, and strength at end of treatment, 6 weeks(no benefit demonstrated). Patients with chronicstroke.

Progressive-resistive robotic training versussensorimotor training, level I (RCT): Grade C+for decrease of spasticity at end of treatment, 6weeks (clinically important benefit demonstratedwithout statistical significance); grade C for motorfunction (Fugl-Meyer upper extremity and MSSfor shoulder and elbow and wrist and hand), andmotor power for shoulder and elbow at end oftreatment, 6 weeks (no benefit demonstrated). Pa-tients with chronic stroke.

Music-making training versus control, level I(RCT): Grade C+ for ROM (elbow extension) atend of treatment, 10 weeks (clinically importantbenefit demonstrated without statistical signifi-cance); grade C for ROM (shoulder flexion) at endof treatment, 10 weeks (no benefit demonstrated).Patients with post-acute stroke.

Water-based training versus control, level I(RCT): Grade A for hip and knee extensorsstrength (affected side) at end of treatment, 8weeks (clinically important benefit demon-strated); grade C+ for cardiopulmonary function(VO

2 max) at end of treatment, 8 weeks, muscle

power at end of treatment, 8 weeks, and gaitspeed at end of treatment, 8 weeks (clinicallyimportant benefit demonstrated without statisti-cal significance); grade C for hip and knee exten-sors strength (unaffected side) at end of treat-ment, 8 weeks (no benefit demonstrated); gradeD for balance at end of treatment, 8 weeks (nobenefit demonstrated but favoring control). Pa-tients with chronic stroke.

Agility exercise versus stretching/weight-shifting exercise, level I (RCT): Grade C+ for stepreaction time at follow-up, 1 month (clinicallyimportant benefit demonstrated without statisticalsignificance); grade C for balance, mobility, bal-

14 TOPICS IN STROKE REHABILITATION/SPRING 2006

ance confidence, and quality of life at end of treat-ment, 10 weeks, and follow-up, 1 month, step reac-tion time at end of treatment, 10 weeks (no benefitdemonstrated). Patients with chronic stroke.

Maximal isokinetic strengthening versus con-trol, level I (RCT): Grade C+ for change instrength at end of treatment, 6 weeks (clinicallyimportant benefit demonstrated without statisticalsignificance); grade C for quality of life and at end

of treatment, 6 weeks (no benefit demonstrated);grade D for level-walking and stair-walking (nobenefit demonstrated but favoring control). Pa-tients with chronic stroke.

Mental imagery versus standard functionaltraining, level I (RCT): Grade A for level of inde-pendence in performing tasks at end of treatment,1 week, 2 weeks, and 3 weeks (clinically importantbenefit demonstrated). Patients with acute stroke.

Summary of trials

Twenty-eight RCTs and one CCT were includedthat evaluated the efficacy of different kinds oftherapeutic exercises in comparison to either an-other type of exercise therapy or to a placebo (N =1,318).99127 Seven subcategories of exercisetherapies were administered: 1) aerobic exercises:cycle ergometer for the lower extremities (n =185) 100,109,110,121,123 and for the upper body (n =40),100 as well as water-based aerobic training (n =12)101; 2) therapies based on theories in neurol-ogy: proprioceptive neuromuscular facilitation(PNF; n = 173),102,115 the Bobath approach (n =40),119 and approaches based on neurophysi-ological and developmental theories (n =47)106,116; 3) exercises for the lower extremity:progressive resistance exercises (n = 230),107,110,118

active exercises (also for the trunk; n = 77),107 andpassive ROM (n = 20)110; 4) exercises for theupper extremity: gross and fine movement exer-cises and muscle strengthening (n = 92), 101,120,127

robot-aided therapy with goal-directed move-ments (n = 97),99,105,116,122,125,126 active exerciseswith a skateboard and overhead pulley and pas-sive ROM exercises for the shoulder (n = 28),111

resisted extension of fingers, ballistic extensionand resisted grasp for the hand (n = 20)121; 5)balance exercises (n = 88)117,119; 6) functionaltraining (n = 106)114,127; 7) mixture of exercisetherapies: ROM and flexibility exercises for theupper and lower extremities and trunk, strength-ening resistive exercises using PNF patterns forupper and lower extremities or theraband exer-cises, balance training, upper extremity func-tional use, and progressive walking program/ex-ercise on a bicycle ergometer (n = 112).103,104 The

length of each exercise session ranged from 20minutes to 2 hours per day, and the treatmentprograms durations ranged from 2 sessions perweek to daily for 2 weeks to 6 months or untildischarge. In Duncan et al.s study,103 patientswere instructed to continue exercising indepen-dently for an additional 4 weeks, while inLanghammers studies, patients continued treat-ment as outpatients112,113 (TE-Appendix 3A).

In total, 31 studies were excluded for the follow-ing reasons: there was no control group in 10 of thestudies,128137 healthy subjects were used in threestudies,138140 two studies had insufficient statisticaldata,141,142 two studies were not specific to strokepatients,143,144 and another study lacked an interven-tion.145 The remaining studies were excluded forvarious other reasons146158 (TE-Table 3).

Efficacy

Aerobic training for patients in the subacuteand chronic phases of stroke recovery versus acontrol group (3 RCTs, n = 145)108,121,123 resulted inclinically important benefits for cardiopulmonaryfunction (expiration per minute [VE]), musclepower according to the maximal workload, andfunctional status (walking measured through theAdjusted Activity Score [AAS]) at the end of 10weeks of treatment (18%, 35%, and 41% RD,respectively). These outcomes were also statisti-cally significant (TE-Figure 1A, Table 4A). Clini-cally important benefits were demonstrated with-out statistical significance for gait speed at the endof 10 weeks of treatment (33% RD; TE-Figure 1A,Table 4A). No benefit was demonstrated for car-diopulmonary function as measured (maximal

Clinical Practice Guidelines 15

heart rate, VO2 max, VCO

2 max) and motor func-

tion (Fugl-Meyer index) at the end of 10 weeks oftreatment (TE-Figure 1A, Table 4A) and for func-tional status (FAI total, social outings, walkingoutside) at 6 months (TE-Figure 1B, Table 4B).Results favored the control group for functionalstatus during light housework activities at the endof 6 months of treatment (TE-Figure 1B, Table4C). No other outcomes were measured.

For individualized aerobic program trainingversus control in patients during the subacutephase of stroke (one RCT, n = 90),109 statisticallysignificant and clinically important benefits werefound for physical fitness (highest test stage com-pleted of the stress test and maximal workload)and mobility (stair climbing) at the end of 8 weeksof treatment (33%, 36%, and 130% RD, respec-tively; TE-Figure 2, Table 5). Clinically importantbenefits without statistical significance were dem-onstrated for mobility (walking distance) at theend of 8 weeks of treatment (28% RD; TE-Figure2, Table 5). No benefit was demonstrated forcardiovascular function (maximal heart rate, de-crease of resting heart rate, decrease of restingsystolic and diastolic blood pressure), gait speed,and functional status (FIM; TE-Figure 2, Table 5)at the end of 8 weeks of treatment. No otheroutcomes were measured.

For PNF versus standard customary muscletraining of the upper extremity in post-acutestroke patients (two RCTs, n = 173),102,115 nobenefit was found for functional status (BarthelIndex) and upper extremity muscle strength(manual muscle test) at the end of 6 weeks oftreatment (TE-Figure 3, Table 6A). Results fa-vored the control group for mobility number(number of patients independent in walking) atthe end of 2, 4, and 6 weeks of treatment (TE-Figure 3, Table 6B) and for ROM of the wristnumber (number of patients with limited ROM) atthe end of 6 weeks of treatment (TE-Figure 3,Table 6B). A clinically important benefit withoutstatistical significance was found that favored con-trol for ROM of the ankle (number of patients withlimited ROM; TE-Figure 3, Table 6B). No otheroutcomes were measured.

For PNF versus the Bobath approach in sub-acute stroke patients (one RCT, n = 131), 102 no

benefits were found for mobility (number of pa-tients independent in walking) at the end of 2, 4,and 6 weeks of treatment and ROM of the wristand ankle the at end of 6 weeks of treatment (TE-Figure 4, Table 7). No other outcomes weremeasured.

For the Bobath approach versus standard cus-tomary muscle training in acute and subacutestroke patients (three RCTs, n = 212),102,112,113 clini-cally important benefits without statistical signifi-cance were found for motor function (SMES upperextremity) at the 4-year follow-up (16% RD), qual-ity of life (NHPglobal) at the 1-year and 4-yearfollow-ups (26% and 22% RD, respectively), andquality of life (NHP loss of energy) at the end of 3months of treatment (23% RD; TE-Figure 5,Table 8). No benefits were demonstrated for mo-bility (number of patients independent in walking)at the end of 2, 4, and 6 weeks of treatments,motor function (SMES lower extremity and trunk,balance and gait, and Motor Assessment Scale) atthe end of 3 months of treatment and at the 1- and4-year follow-up, motor function (SMES upperextremity) at the end of 3 months of treatment andat the 1-year follow-up, and functional status(Barthel Index) at 4 years follow-up (TE-Figure 5,Table 8). There was no benefit demonstrated, butresults favored the control therapy for ROM of thewrist and ankle at the end of 6 weeks of treatmentand functional status (Barthel Index) at the end of3 months of treatment and at the 1-year follow-up(TE-Figure 5, Table 8). A clinically important ben-efit was found without statistical significance favor-ing the control therapy for pain relief (NHPpain) atthe end of 3 months of treatment (TE-Figure 5,Table 8). No other outcomes were measured.

For Bobath approach training in subacutestroke patients versus control (one RCT, n =40),119 a clinically important benefit without statis-tical significance was found for standing balance atthe 2 and 12 week follow-up (22%35% RD). Nobenefit was demonstrated for sitting balance at endof 4 weeks of treatment. Results favored the con-trol group for sitting balance at the 2-week follow-up and standing balance at the end of 4 weeks oftreatment. A clinically important benefit withoutstatistical significance that favored control wasdemonstrated for sitting balance at the 12-week

16 TOPICS IN STROKE REHABILITATION/SPRING 2006

follow-up (TE-Figure 6, Table 9). No other out-comes were measured.

For progressive resistance training versus ac-tive training of the lower extremity in post-acute stroke patients (one RCT, n = 77),107 aclinically important benefit with statistical signifi-cance was found for functional status (number ofpatients improved in more than two activities ofdaily living [ADL] items; 34% RD) at end of treat-ment, 1 month (TE-Figure 7, Table 10). No otheroutcomes were measured.

For active training of the lower extremity inpost-acute stroke patients versus control (noexercise) (one RCT, n = 77), 107 no benefit wasdemonstrated, but results favored control for func-tional status (number of patients improved inmore than two ADL items) without statistical sig-nificance at end of 1 month of treatment (TE-Figure 8, Table 11). No other outcomes weremeasured.

For progressive resistance training of thelower extremity in post-acute stroke patientsversus control (one RCT, n = 77),107 clinicallyimportant benefits without statistical significancewere found for functional status (number of pa-tients improved in more than two ADL items) atthe end of 1 month of treatment (26% RD; TE-Figure 9, Table 12). No other outcomes weremeasured.

For progressive resistance in subacute strokepatients versus no resistance training (one RCT,n = 133),118 no benefit was shown for motor recov-ery (Chedoke-McMaster Stroke Assessment[CMSA]: walking and gross motor function indexsection) at the end 4 weeks and 8 weeks of treat-ment and at the 6-month follow-up. No benefitwas demonstrated, but results favored control forgait endurance (2-minute walking test) at the endof 4 and 8 weeks of treatment and at the 6-monthfollow-up (TE-Figure 10, Table 13). No otheroutcomes were measured.

For functional task training of the upper ex-tremity versus strength training in subacutestroke patients (one RCT, n = 60),127 clinicallyimportant benefits without statistical significancewere shown favoring functional task training forfunctional status (FIM self-care and FIM mobility;24% and 17% RD, respectively), upper extremity

isometric torque (62% RD), grip strength (317%RD), and palmar pinch (119% RD) at the 6.5- to 8-month follow-up and lateral pinch at the end of 4weeks of treatment at the 9-month follow-up (39%and 47% RD, respectively) (TE-Figure 11, Table14). A clinically important benefit without statisti-cal significance was demonstrated favoringstrength training for grip strength (30% RD) andpalmar pinch (19% RD) at the end of 4 weeks oftreatment (TE-Figure 11, Table 14). No benefitwas demonstrated for upper extremity ROM, painrelief, upper extremity sensory and motor func-tion, and functional status (FTHUE) at the end of 4weeks of treatment and at the 6.5- to 8-monthfollow-up and isometric torque at the end of 4weeks of treatment (TE-Figure 11, Table 14). Noother outcomes were measured.

For strength training in subacute stroke pa-tients versus control (one RCT, n = 60),127 clini-cally important benefits were found with statisti-cal significance for upper extremity isometrictorque (34% RD; TE-Figure 12, Table 15) at theend of 4 weeks of treatment. A clinically impor-tant benefit was demonstrated without statisticalsignificance for motor function of the upper ex-tremity (42% RD) and functional status, mea-sured by the FTHUE (17% RD) at the end of 4weeks of treatment, palmar pinch at the end of 4weeks of treatment and at the 6.5- to 8-monthfollow-up (49%54% RD), and grip strength(36% RD) and lateral pinch (30% RD) at the 6.5-to 8-month follow-up (TE-Figure 12, Table 15).No benefit was demonstrated for sensory func-tion of the upper extremity at the end of 4 weeksof treatment and at the 6.5- to 8-month follow-up, functional status (FIM mobility), and painrelief and grip strength at the end of 4 weeks oftreatment (TE-Figure 12, Table 15). Clinicallyimportant benefits favoring control without sta-tistical significance were demonstrated for lateralpinch at the end of 4 weeks of treatment, func-tional status (FIM self-care and FIM mobility),and upper extremity isometric torque at the 6.5-to 8-month follow-up (TE-Figure 12, Table 15).The results favored control for upper extremityROM at the end of 4 weeks of treatment and atthe 6.5- to 8-month follow-up and functionalstatus (FTHUE) at the 6.5- to 8-month follow-up

Clinical Practice Guidelines 17

(TE-Figure 12, Table 15). No other outcomeswere measured.

For combined aerobic and strength versusaerobic training alone in chronic stroke patients(one RCT, n = 40),100 clinically important benefitswith statistical significance were shown for car-diopulmonary function (VO

2 max; 18% RD) and

peak torque of shoulder flexors (47% RD) at theend of 16 weeks of treatment. A clinically impor-tant benefit was demonstrated without statisticalsignificance for peak torque of knee flexors (16%RD) at the end of 16 weeks of treatment. Nobenefit was demonstrated for peak torque ofshoulder extensors and peak torque of knee exten-sors at the end of 16 weeks of treatment (TE-Figure 13, Table 16). No other outcomes weremeasured.

For Kinetron training of the lower extremityin stroke patients during the post-acute phaseof recovery versus control (no Kinetron; oneRCT, n = 20),106 no benefit was found, but resultsfavored control for mobility (FunctionalAmbulation Profile [FAP] test) at the end of 5weeks of treatment (TE-Figure 14, Table 17). Noother outcomes were measured.

For home-based exercise training in post-acute stroke patients versus control (two RCTs,n = 112),103,104 clinically important benefits withstatistical significance were shown for change ingait speed (48% RD), gait endurance (change in 6-minute walk test, feet; 59% RD), torque (change inknee isometric extensors; 61% RD), endurance(change in duration of bicycle exercises; 164%RD), and cardiopulmonary function (change inpeak VO

2; 185% RD) at the end of 12 weeks of

treatment (TE-Figure 15, Table 18). A clinicallyimportant benefit was demonstrated without sta-tistical significance for motor function (change inFugl-Meyer lower extremity; 16.7% RD), changein gait speed (94% RD), gait endurance (change in6-minute walk test, feet; 15% RD), functional sta-tus (physical function index; 20% RD), andstrength (change in grip strength; 17% RD) at theend of 12 weeks of treatment (TE-Figure 15,Table 18). No benefit was demonstrated for motorfunction (change in Fugl-Meyer upper extremityand lower extremity), balance (Berg balance andchange in Berg balance), or functional status (In-

strumental ADL and Barthel ADL Index) at the endof 12 weeks of treatment (TE-Figure 15, Table18). A clinically important benefit without statisti-cal significance was shown favoring control forbalance (functional reach) at the end of 12 weeksof treatment (TE-Figure 15, Table 18). No benefitwas demonstrated, but results favored control fortorque (change in ankle isometric dorsiflexors) atthe end of 12 weeks of treatment (TE-Figure 15,Table 18). No other outcomes were measured.

For skateboard use versus overhead pulleytraining of the affected shoulder in subacutestroke patients (one RCT, n = 28),111 a clinicallyimportant benefit without statistical significancewas found for pain relief (number of patients with-out shoulder pain) at the end of 810 weeks oftreatment (50% RD; TE-Figure 16, Table 19). Noother outcomes were measured.

Comparing the use of an overhead pulleyversus control (passive ROM training for shoul-der) for subacute stroke patients (one RCT, n =28),111 no benefit was found but results favoredcontrol for pain relief (number of patients withoutpain) at the end of 810 weeks of treatment (TE-Figure 17, Table 20). No other outcomes weremeasured.

For passive ROM training of the affectedshoulder versus skateboard use in subacutestroke patients (one RCT, n = 28),111 no benefit wasshown for pain relief (number of patients withoutpain) at the end of 810 weeks of treatment (TE-Figure 18, Table 21). No other outcomes weremeasured.

For resisted extension versus ballistic exten-sion training of the hand in subacute and post-acute stroke patients (one RCT, n = 20),121 a clini-cally important benefit without statisticalsignificance was found for ROM change in fingerROM at the end of 2 weeks of treatment (42% RD).No benefit was demonstrated for motor function(change in grasp/release and change in status-tap-ping; TE-Figure 19, Table 22). No other out-comes were measured.

For resisted extension exercises versus re-sisted grasp training of the hand in subacute andpost-acute stroke patients (one RCT, n = 20),121 aclinically important benefit with statistical signifi-cance was found for motor function (change in

18 TOPICS IN STROKE REHABILITATION/SPRING 2006

tapping) at the end of 2 weeks of treatment (115%RD). A clinically important benefit was demon-strated without statistical significance for ROM(change in finger ROM) at the end of 2 weeks oftreatment (50% RD). No benefit was demonstratedfor motor function (change in grasp/release; TE-Figure 20, Table 23). No other outcomes weremeasured.

For resisted extension training of the hand insubacute and post-acute stroke patients versuscontrol (one RCT, n = 20),121 a clinically importantbenefit with statistical significance was shown formotor function (change in tapping) and ROM(change in finger ROM) at the end of 2 weeks oftreatment (118% and 122% RD, respectively). Nobenefit was demonstrated, but results favored con-trol for motor function (change in grasp/release) atthe end of 2 weeks of treatment (TE-Figure 21,Table 24). No other outcomes were measured.

For ballistic extension versus resisted grasptraining of the hand in subacute and post-acutestroke patients (one RCT, n = 20),121 a clinicallyimportant benefit with statistical significance wasfound for motor function (change in tapping) atthe end of 2 weeks of treatment (108% RD). Aclinically important benefit was demonstratedwithout statistical significance for ROM (change infinger ROM) at the end of 2 weeks of treatment(91%). No benefit was demonstrated for motorfunction (change in grasp/release) and ROM(change in finger ROM) at the end of 2 weeks oftreatment (TE-Figure 22, Table 25). No otheroutcomes were measured.

For ballistic extension training of the hand insubacute and post-acute stroke patients versuscontrol (one RCT, n = 20),121 a clinically importantbenefit without statistical significance was shownfor motor function (change in tapping) and ROM(change in finger ROM) at the end of 2 weeks oftreatment (112% and 97% RD, respectively). Nobenefit was demonstrated for ROM (change infinger ROM) at the end of 2 weeks of treatment,and no benefit was demonstrated but results fa-vored control for motor function (change in grasp/release) at the end of 2 weeks of treatment (TE-Figure 23, Table 26). No other outcomes weremeasured.

For resisted grasp training of the hand insubacute and post-acute stroke patients versus

control (one RCT, n = 20), 121 a clinically impor-tant benefit without statistical significance wasfound for ROM (change in finger ROM) at the endof 2 weeks of treatment (84% RD). No benefit wasdemonstrated for motor function (change in tap-ping) at the end of 2 weeks of treatment. Nobenefit was demonstrated, but results favored con-trol for motor function (change in tapping andchange in grasp/release) and ROM (change in fin-ger ROM) at the end of 2 weeks of treatment (TE-Figure 24, Table 27). No other outcomes weremeasured.

For robot-aided training in subacute-chronicstroke patients versus no robot-aided training(two RCTs and one CCT, n = 88),99,125,126 clinicallyimportant benefits with statistical significancewere shown for motor power for shoulder andelbow at the end of 5 weeks of treatment (124%RD), change in motor power upper extremity atthe end of 6 weeks treatment (15% RD), motorfunction (Fugl-Meyer for shoulder and elbow andMSS for shoulder and elbow) at the end of 5 weeksof treatment (18% and 43% RD, respectively), mo-tor function (MSS for wrist and hand) at the end of5 weeks of treatment (40% RD), and motor func-tion (change in MSS for shoulder and elbow) at theend of 6 weeks of treatment and at 3 years follow-up (32.5% and 30.5% RD, respectively; TE-Figure25A, Table 28A). A clinically important benefitwith statistical significance was demonstrated formotor function (MSS for upper extremity) at theend of 6 weeks of treatment (64% RD; TE-Figure25A, Table 28A). Clinically important benefitswere demonstrated without statistical significancefor change in motor power for shoulder and elbowat 3 years follow-up (20% RD), motor function(Fugl-Meyer scale for upper extremity; 26% RD) atthe end of 6 weeks of treatment, and motor powerfor upper extremity at the end of 6 weeks of treat-ment (37% RD; TE-Figure 25A&B, Table 28A&B). No benefit was demonstrated for motorfunction (change in Fugl-Meyer for shoulder, el-bow, and coordination) at the end of 6 weeks oftreatment, functional status (FIM for upper ex-tremity) at the end of 5 weeks and 6 weeks oftreatment, and motor function (change in Fugl-Meyer for wrist and hand and change in MSSfor wrist and hand) at the end of 6 weeks oftreatment and at 3 years follow-up (TE-Figure

Clinical Practice Guidelines 19

25A&B, Table 28A&B). No benefit was demon-strated, but results favored control for motor func-tion (change in Fugl-Meyer for shoulder, elbow,and coordination; TE-Figure 25A, Table 28A).No other outcomes were measured.

For robot-assisted versus neurodevelopmen-tal (NDT) training in chronic stroke patients(one RCT, n = 27), 116 a clinically important benefitwith statistical significance was found for strength(change in elbow extensors, shoulder internal rota-tors, abductors, adductors, and flexors strength[%]) and functional reach (change in forward me-dial, forward, forward lateral, and lateral reachextent) at the end of 2 months of treatment (270%,99%, 134%, 107%, 160%, 270%, 175%, 229%,and 612% RD, respectively; TE-Figure 26A&B,Table 29A&B). A clinically important benefit wasdemonstrated without statistical significance forstrength (change