Methods Randomized Trial Wst

7/27/2019 Methods Randomized Trial Wst

1/16

http://nnr.sagepub.com

Neurorehabilitation and Neural Repair

DOI: 10.1177/08884390032555082003; 17; 153Neurorehabil Neural RepairDudley, Robert Elashoff, Lisa Fugate, Susan Harkema, Michael Saulino and Michael Scott

Bruce H. Dobkin, David Apple, Hugues Barbeau, Michele Basso, Andrea Behrman, Dan Deforge, John Ditunno, Gfor Walking During Inpatient Rehabilitation after Incomplete Traumatic Spinal Cord Injury

ethods for a Randomized Trial of Weight-Supported Treadmill Training Versus Conventional Train

http://nnr.sagepub.com/cgi/content/abstract/17/3/153The online version of this article can be found at:

Published by:

http://www.sagepublications.com

On behalf of:

American Society of Neurorehabilitation

can be found at:Neurorehabilitation and Neural RepairAdditional services and information for

http://nnr.sagepub.com/cgi/alertsEmail Alerts:

http://nnr.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://nnr.sagepub.com/cgi/content/refs/17/3/153Citations

at Scott Memorial Library @ Thomas Jefferson University on January 26, 2009http://nnr.sagepub.comDownloaded from
http://www.asnr.com/http://nnr.sagepub.com/cgi/alertshttp://nnr.sagepub.com/cgi/alertshttp://nnr.sagepub.com/subscriptionshttp://nnr.sagepub.com/subscriptionshttp://nnr.sagepub.com/subscriptionshttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsPermissions.navhttp://nnr.sagepub.com/cgi/content/refs/17/3/153http://nnr.sagepub.com/http://nnr.sagepub.com/http://nnr.sagepub.com/http://nnr.sagepub.com/http://nnr.sagepub.com/cgi/content/refs/17/3/153http://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsReprints.navhttp://nnr.sagepub.com/subscriptionshttp://nnr.sagepub.com/cgi/alertshttp://www.asnr.com/


2/16

Copyright 2003 The American Society of Neurorehabilitation 153

The authors describe the rationale and methodology forthe first prospective, multicenter, randomized clinicaltrial (RCT) of a task-oriented walking intervention forsubjects during early rehabilitation for an acute trau-matic spinal cord injury (SCI). The experimental strategy,body weightsupported treadmill training (BWSTT),allows physical therapists to systematically train patientsto walk on a treadmill at increasing speeds typical ofcommunity ambulation with increasing weight bearing.The therapists provide verbal and tactile cues to facilitatethe kinematic, kinetic, and temporal features of walking.Subjects were randomly assigned to a conventional ther-apy program for mobility versus the same intensity andduration of a combination of BWSTT and over-groundlocomotor retraining. Subjects had an incomplete SCI

(American Spinal Injury Association grades B, C, and D)from C-4 to T-10 (upper motoneuron group) or from T-11to L-3 (lower motoneuron group). Within 8 weeks of aSCI, 146 subjects were entered for 12 weeks of interven-tion. The 2 single-blinded primary outcome measures arethe level of independence for ambulation and, for thosewho are able to walk, the maximal speed for walking 50feet, tested 6 and 12 months after randomization. Thetrials methodology offers a model for the feasibility of

translating neuroscientific experiments into a RCT todevelop evidence-based rehabilitation practices.

Key Words:Motor learning-Locomotor training-Neurologic rehabilitation-Spinal cord

Ambulation is compromised in most of themore than 10,000 yearly survivors of a trau-matic spinal cord injury (SCI) and for250,000 people in the United States with chronicSCI.1 Patients with complete sensorimotor loss(graded by the American Spinal Injury Association

scale [ASIA] as A; see Table 12

) and those gradedASIA B within 1 week to 1 month of the SCI recov-er walking in no more than 10% to 15% of cases.Most subjects who have had enough return ofmotor control to regain the ability to walk will doso at greater than normal energy cost. Outcomestudies of ambulation vary with the method used toclassify impairment, the time of assessment afterSCI, and the measures employed. Previous obser-

vational studies of the recovery of walking afterSCI, such as those from the American ModelSystems program,3 have not prospectively assessedrelationships between functional recovery, rate of

recovery, walking speed and endurance, and qual-ity of life in patients who were admitted for inpa-tient rehabilitation.

During inpatient rehabilitation soon after SCI,physical therapy ordinarily proceeds beyond sup-ported standing only if the patients legs are bracedto lock them in extension or if the patient has thestrength and balance necessary to maintain the legs

Methods for a Randomized Trial ofWeight-Supported Treadmill Training

versus Conventional Training for Walking

during Inpatient Rehabilitation afterIncomplete Traumatic Spinal Cord Injury

Bruce H. Dobkin, David Apple, Hugues Barbeau, Michele Basso,Andrea Behrman, Dan Deforge, John Ditunno, Gary Dudley, Robert Elashoff,

Lisa Fugate, Susan Harkema, Michael Saulino, and Michael Scott,for the Spinal Cord Injury Locomotor Trial (SCILT) Group

From the Department of Neurology, Reed Neurologic ResearchCenter, University of California, Los Angeles.

Address correspondence to Bruce H. Dobkin, M.D., Departmentof Neurology, Neurologic Rehabilitation and Research Program,University of California Los Angeles, Reed Neurologic Research

Center, 710 Westwood Plaza, Los Angeles, CA 90095. E-mail:[email protected].

Dobkin BH, Apple D, Barbeau H, Basso M, Behrman A, DeforgeD, Ditunno J, Dudley G, Elashoff R, Fugate L, Harkema S,Saulino M, Scott M, for the Spinal Cord Injury Locomotor Trial(SCILT) Group. Methods for a randomized trial of weight-sup-ported treadmill training versus conventional training for walk-ing during inpatient rehabilitation after incomplete traumaticspinal cord injury. Neurorehabil Neural Repair 2003;17:153-167.

DOI: 10.1177/0888439003255508

http://nnr.sagepub.com/http://nnr.sagepub.com/http://nnr.sagepub.com/http://nnr.sagepub.com/


3/16

in extension during weight bearing and to flex atthe hips and knees when taking steps. If the armsare capable of partially supporting the persons

weight during stepping in parallel bars or in awalker without inducing great energy cost, gaittraining proceeds.

The Spinal Cord Injury Locomotor Trial (SCILT) isdesigned to compare conventional inpatient and

outpatient physical therapy to an intervention oftencalled body weightsupported treadmill training(BWSTT).4 This technique for locomotor retrain-

ing partially supports the weight of a patient by

a parachute-type harness attached at the shouldersto an overhead lift. Initial weight support preventsthe paraparetic legs from buckling at the knees.The lift allows vertical displacement during step-ping and supports up to 50% of the subjects

weight. Therapists then systematically train patientsto walk on the treadmill at increasingly fasterspeeds and reduce the amount of weight support

when feasible. The therapists aim to optimize thekinematic, kinetic, and temporal components ofgait that are tied to the stance and swing phases of

walking. This training aims to facilitate walking-related sensory inputs for a rhythmical, reciprocalgait pattern, without the requirements of good pos-tural stability and full weight bearing. The trainingalso carries over facilitative cues and an emphasison sensory inputs relevant to stance and swing for

walking over ground and in the community.Several quasi-experimental studies in SCI sug-

gested that BWSTT may increase the likelihood thatsubjects graded ASIA B, C, and D with upper

motoneuron (UMN) injuries will learn to walk overground.5-9 Experimental neuroscientific studies sug-gested that improved motor output and locomotionafter SCI may be inducible by mechanisms of neu-roplasticity that are influenced by task-orientedsensorimotor training and paradigms for motorlearning. These include 1) sprouting from primarysensory neurons and from residual descendinginputs to the lumbosacral motor pools; 2) modifi-cation of spinal gray matter synaptic transmis-sion10,11; 3) conduction through demyelinated butintact axons12,13; 4) locomotor movement patternsfacilitated by lumbar spinal cord central pattern

generators and related circuits for stepping10,14-20; 5)enhancement of activity in residual, subclinicaldescending pathways that must be trained to playa greater role in posture and gait21,22; and 6)changes in residual cortical and subcortical neu-ronal assemblies that represent the kinematics andforces associated with walking.23-25

BWSTT does not rely only on theories about cen-tral pattern generators, especially not in incompletesubjects who have some lower extremity motorcontrol. SCILT included subjects who had lowermotoneuron (LMN) injury because 20% of people

with traumatic SCI have such lesions. The LMNgroup may not have access to the lumbosacral gen-erators if damaged by an injury near the T-12 ver-tebral body. These subjects may still make use ofsegmental sensory inputs and reorganization with-in the cord above the lesion, as well as withinbrainstem and cortical sensorimotor networks.

SCILT, then, draws its face validity from a gener-al hypothesis about motor skills relearning inpatients with a moderate to profound loss ofascending and descending spinal pathways. By

assisting patients to step so that proprioceptive andcutaneous sensory inputs to the spinal cord andbrain are more typical of ordinary walking than

what patients with paraparesis achieve during con-ventional training, and by intensive practice of suchstepping on a treadmill, mechanisms of activity-dependent plasticity may increase rehabilitativegains.1

Progress in acute protection of spinal cord neu-rons and tracts and in the regeneration of axonsand neuronal connections with, for example, neu-rotrophic factors, immune blocking substances, cellimplants, and neural bridges across the level of

injury offers hope for biological interventions forSCI in patients.26 Well-defined training strategies

will also be necessary to optimize the functionalutility of such interventions. BWSTT may be amethod to enhance the induction of activity-dependent plasticity within new, as well as resid-ual, pathways and networks.27-30

B. H. Dobkin et al.

154 Neurorehabilitation and Neural Repair 17(3); 2003

Table 1. American Spinal Injury Association (ASIA)Impairment Scalea

AComplete: no sensory or motor function below thelesion, including sacral segments S-4 and S-5.

BIncomplete: sensory but not motor function ispreserved below the neurological level andincludes S-4 and S-5.

CIncomplete: motor function is preserved below thelevel and more than half of key muscles below thelevel are graded


4/16

TRIAL DESIGN AND METHOD

Overview

This prospective, single-blinded randomized clin-ical trial (RCT) included 2 arms: an experimentalgroup that received BWSTT and over-ground gaittraining and a control group that received conven-

tional standing and mobility training for the sameamount of practice time. The protocol simulatedcurrent rehabilitation practices for acute traumaticSCI rehabilitation in terms of the amount of thera-py related to mobility training, usual admission anddischarge practices for facilities, and typical periodsof inpatient and outpatient care. The investigators

were concerned that the ordinary amount of mobil-ity retraining provided during inpatient and earlyoutpatient rehabilitation care may be less than opti-mal but decided that the study should proceed

within common practice. Patients with incompleteSCI were stratified by severity using the ASIAimpairment scale (Table 1) and by UMN or LMNinjury.

The study addresses the following primaryhypotheses:

1. Subjects classified as ASIA B or C at entry whoreceived BWSTT recover supervised to independ-ent walking with a reciprocal gait pattern for 150feet (a Functional Independence Measure [FIM]locomotor score 5) significantly more often com-pared to those given only conventional therapy.

2. Subjects classified as ASIA D at entry who receivedBWSTT will walk 50 feet over ground significantlyfaster than conventionally treated subjects.

The secondary hypotheses are the following:

3. Subjects graded ASIA B or C on entry who walkwith moderate assistance or less help at the time ofoutcome assessments will walk faster if theyreceived BWSTT; more ASIA D subjects will reacha locomotor score of 5 if they received BWSTT.

4. Subjects who received BWSTT and walk withmoderate assistance or less help at the time of out-come assessments have greater endurance andwalk farther in 6 minutes compared to patientstreated with conventional physical therapy.

5. Subjects who received BWSTT will recover a func-tional gait significantly sooner after the SCI com-

pared to those who received conventional thera-pies alone.

6. Subjects who received BWSTT will be more inde-pendent in personal and community activities andwill perceive their quality of life as being better.

7. Earlier step training permitted by BWSTT willreduce medical complications related to immobility.

8. Subjects who received BWSTT will have better sit-ting and standing balance as measured by the BergScale.

9. Subjects who received BWSTT will have less spas-ticity of the lower extremities, as measured by theAshworth Scale, and fewer reflexive spasms duringthe time of therapy.

Inclusion and Exclusion Criteria

Subjects were rated by the ASIA at the time ofrandomization. Subjects graded ASIA B had to havea neurological level below C-6, so they could usetheir upper extremities to manage assistive devices.Subjects graded ASIA C and D could have a levelbelow C-4 as long as they fully extended theirelbows against gravity. The cervical to T-10/11-level subjects were designated as the UMN group,and the T-11 to L-3 subjects had conus/caudaequina lesions and were designated the LMNgroup. Subjects with a LMN lesion and absentsacral sensation were eligible and designated forthe SCILT trial as ASIA B if any lumbar sensation

was present, even without sacral preservation ofsensation, and as ASIA C or D if motor function

was present at L-3 or above. This classification forsubjects with a LMN lesion differs from the usual

ASIA B definition. Early after a SCI, the clinical dis-tinction between a UMN and LMN designation maybe equivocal, especially in subjects who have lowthoracic sensorimotor findings. An absence of deeptendon reflexes in the lower extremities, absenceof a bulbocavernous reflex, hypotonicity of thelegs, and a neurogenic bladder without hyper-reflexia of the detrusor or detrusor-sphincterdyssynergia point to a LMN classification. When

doubts persisted, a significantly low amplitudecompound action potential of the posterior tibial orperoneal nerves from 2 to 8 weeks after the SCI

was an optional test to demonstrate the presence orabsence of LMN axonal involvement. Subjects mis-classified as UMN or LMN at the time of random-ization were reclassified by the Clinical Unit (CU)before the outcomes measures taken at 3 months,after documenting the basis for the change with theprincipal investigator (PI) and the StatisticalCoordinating Unit (SCU) at the University ofCalifornia, Los Angeles (UCLA).

No inclusion/exclusion criteria were based ongender, childbearing potential, or racial or ethnicorigin. Children and adolescents younger than age 16

were excluded because they represent a small per-centage of cases of SCI, may not be mature enoughto work within the training protocol, and maychange in their long bone length over the course ofserial assessments, which may confound the meas-ure of walking speed. Subjects older than 70 years

RCT for Walking after SCI

Neurorehabilitation and Neural Repair 17(3); 2003 155



5/16

were excluded because they too are a small per-centage of patients with uncomplicated traumaticSCI. These patients may have had cervical spondy-losis that interfered with motor control prior to theacute injury. Premorbid medical problems and dis-abilities are also more often present, which mayinterfere with the training protocol and the inter-pretation of the functional outcome measures.

INCLUSION CRITERIA

1. Males and females between ages 16 and 70.2. Traumatic acute SCI within 56 days of injury.3. Incomplete lesion (ASIA B, C, or D at time of ran-

domization) from below C-4 on at least one side ofthe body to no lower than L-3 on either side of thebody.

4. Unable to ambulate over ground at time of ran-domization without at least moderate assistance(scored as a 3 or less by FIM locomotor criterion).

5. Able to offer at least 3/5 strength in the elbowextensors at time of randomization.

6. No clinically significant cognitive impairment

(Mini-Mental Status Score 26 with figure drawingand sentence writing task scored as normal if sub-ject cannot hold a pencil).

7. Signed consent form approved by the InstitutionalReview Board of the site.

EXCLUSION CRITERIA

1. Symptomatic fall in blood pressure or fall greaterthan 30 mm Hg when upright in the body weightsupport apparatus. If this could not be corrected inthe days prior to randomization, patient wasexcluded.

2. Subject with a halo or other cervical brace or tho-racolumbar support orthosis (TLSO) if the treating

physician found BWSTT and wearing a harness tobe medically contraindicated.

3. Contraindication to weight bearing on lowerextremities (pelvic or leg fracture, chronic jointpain).

4. Pressure sore with any skin breakdown (beyondstage 2, using Model Systems criteria), where aharness or treadmill training or standing could, inthe physicians judgment, negatively affect healing.

5. Any debilitating disease prior to the acute SCI thatcaused exercise intolerance or more than minimal-ly limited mobility-related self-care and instrumen-tal activities of daily living. This included exertion-al angina, heart failure, and clinically significantpulmonary disease (class 3 or more by American

Heart Association standard); neurologic diseasessuch as Parkinsons, peripheral sensory or motorneuropathy, prior traumatic brain injury, andstroke; a malignancy likely to cause death within 2years; alcoholic liver disease; and chronic alcoholor drug abuse.

6. Subject must use an antispasticity medication at thetime of entry, with the exception of use limited tobedtime to prevent spasms that interfere withsleep.

7. Premorbid, ongoing major depression or psy-chosis; suicide attempt that caused the SCI.

8. Subject unlikely to complete the intervention orreturn for follow-up due to distance from home toassessment site. Subjects could be assessed at oneof the other sites at 3, 6, and 12 months if necessary.

9. Participation in another rehabilitation or pharma-cologic study that continued beyond the start ofthe trial.

Sample Size and Power

The investigators at each CU reviewed their insti-tutional databases for the previous 2 years (Tables2 and 3). Only 15% of their patients classified as

ASIA B at the time of admission for rehabilitationwere able to walk 150 feet at a supervised or bet-ter level of function at discharge. About 40% oftheir ASIA C and 75% of their ASIA D subjectsachieved a FIM locomotor score of 5 (Table 2).Because of the divergence between the likelihood

of scoring 5 between the ASIA B and C subjectscompared to the ASIA D subjects, the level of inde-pendence and the walking speed, respectively,

were chosen as separate primary endpoints forSCILT. The primary outcomes are to be assessedseparately for each of the B, C, and D ASIA stratifi-cations, rather than combining the B and C groups,because all have different likelihoods of achievinga score of 5 or greater on the FIM locomotor scale.Walking speed reflects residual supraspinal input

measured by lower extremity strength, as well asthe energy cost of locomotion, and predicts who

will evolve into a functional ambulator. Waters andcolleagues31,32 assessed a group of subjects withquadriparesis and paraparesis after SCI, most of

whom ambulated with orthoses and assistivedevices. The evaluation included gait velocity, oxy-gen cost per meter walked, and peak axial load ontheir canes. The ability to walk correlated withlower limb strength. Mean walking speed was 40 12 m/min for 34 ambulatory subjects who used areciprocal gait without knee-ankle-foot orthoses(KAFOs) (but with a variety of upper extremityassistive devices), 30 m/min for subjects with para-plegia who needed AFOs, and 26 m/min in sub-

jects with low lumbar SCI who employed bilateralAFOs. Heart rates averaged 130 beats/min, sug-gesting a significant energy cost to walk 50 m.Normal walking speed in these comparisons was80 m/min. Mean walking speed was 31 12 m/minin 10 subjects with SCI classified as ASIA D from 6to 24 months after injury when tested at UCLA. Theover-ground speeds achieved in these studies wereused for the statistical design.

B. H. Dobkin et al.




6/16

For the ASIA B and C subject trial, the sample

size was derived from the data obtained from theCUs (Table 3). The study hypothesis is that theprobability is higher that BWSTT compared to con-

ventional therapy leads to a FIM score 5. Theprobability of a type 1 error was set at 0.05 signifi-cance level. The test statistic for the alternativehypothesis was based on the weighted log oddsratio with standard error estimated by the bootstrapmethod. We chose a power of 0.92 with alpha =

0.05 and a two-sided test for a total sample size of50 ASIA B and 80 ASIA C subjects. To detect greaterthan a 20% difference in walking speed in the ASIAD subjects, a sample size of 80 subjects produceda power of 0.84. Thus, the 2 trials required up to210 subjects (or up to 235 with 10% attrition of sub-jects). Recruitment was planned for 2 to 2 years

based on the recent experience of the CUs. It wasanticipated, however, that recruitment of subjects

with UMN lesions would proceed ahead of subjectswith LMN lesions. Based on local communitydemographics and Table 2, we anticipated that 35%of subjects would be underrepresented minoritiesand 25% women.

Entries into the RCT were stopped when theexpected number of UMN subjects had beenentered and the pace of entry for LMN subjects ledthe Safety Committee and investigators to acknowl-edge that SCILT would have to extend entries for a

year beyond the anticipated end of the trial to enter

the planned number of LMN subjects.

Facilities Contacted

The trial was initiated by the NeurologicRehabilitation and Research Program at theUniversity of California, Los Angeles in response toa 1998 National Institutes of Health request for clin-ical trial proposals. Developers of a proposal withthe UCLA group included Dr Barbeau at McGillUniversity and Drs Fugate and Basso at The Ohio

State University. The PI contacted physicians at largeSCI rehabilitation programs, met with interestedgroups at the ASIA national meeting and the annu-al meeting of the American Society for NeurologicRehabilitation, and sent e-mails to investigators

with relevant interests. To be considered for partic-ipation, sites had to provide data about their inpa-tient SCI services for the past 2 years and demon-strate their ability to enter at least 20 eligible sub-



Table 2. Site Data for Patients with Acute Traumatic Spinal Cord Injury (SCI) in 1997

Ethnicity (%): Total SCIMean Onset Mean White, Hispanic, Patients per

of SCI to Inpatient African American, Year / SCILTRehab Admit Rehab Stay Asian, Eligible Subjects

Clinical Unit (days) (days) Native American per Year

National Uniform Data System, 1995 28 38 31 N/A N/AMagee/Jefferson 26 27 59, 6, 34, 0.5, 0 178 / 26

IRM/McGill-ORH 28 38 93, 0, 3, 1, 3 126 / 23Ohio State 15 29 85, 1, 13, 1, 0 98 / 20Rancho Los Amigos 13 33 22, 47, 25, 5, 0.5 78 / 20Shepherd 20 42 55, 1, 29, 2, 0.5 140 / 32

SCILT, Spinal Cord Injury Locomotor Trial.

Table 3. Rehabilitation Admission ASIA Score and Levelversus Discharge FIM Walking Score for All CUs in 1996and 1997 (SCILT-eligible subjects)

ASIA B ASIA C ASIA D

Level


7/16

jects yearly (Tables 2 and 3). The investigators ateach site agreed not to provide BWSTT outside ofthe trial to any subject who would have been eligi-ble. Investigators at 4 of the interested sites, MageeRehabilitation Center/Jefferson University inPhiladelphia, The Ohio State University in Columbus,Rancho Los Amigos Rehabilitation Center in Los

Angeles County, and Shepherd Rehabili tation

Center in Atlanta, met the requirements anddemonstrated institutional support. By combiningthe rehabilitation services at McGill University andthe University of Ottawa, a fifth site qualified. Thesites in Canada and at The Ohio State Universityhad prior experience in BWSTT, which added totheir value as participants. All centers had achievednational distinction as clinical research centers andas regional SCI centers of excellence for patientcare. Three were long-standing members of the SCIModel Systems Program sponsored by theDepartment of EducationRancho Los Amigos,Magee, and Shepherd.

Table 2 is based on the experience of the sites in1997 on caring for subjects who met the proposedinclusion and exclusion criteria for the study. Thesedata compared favorably with national trends foronset of SCI to time of transfer for admission to arehabilitation facility and for mean inpatient reha-bilitation length of stay, as reported for the 3444acute SCI patients by the 1995 Uniform Data Servicefor Medical Rehabilitation.33 The mean age for trau-matic SCI subjects nationally was 42 years, whichagain was within several years of the mean age ofsubjects admitted to the 5 CUs. Table 3 shows the

distribution of patients at the sites who would havebeen eligible in 1996 and 1997 for SCILT.The UCLA site developed the initial protocol and

data collection and analysis procedures. At severalmeetings of the investigators, the operations manu-al was modified to meet concerns and new infor-mation provided by the CUs. The UCLA site chosenot to participate as a site for subject recruitmentand training so that it could best manage the trialat arms length. The PI did not have access to therandomization procedure or to outcome data byassigned intervention. By remaining blinded, he

was able to advise the site investigators and thera-

pists, the SCU under Robert Elashoff, PhD, andwork with an external Safety and Data MonitoringCommittee.

CU Training

Reported studies using BWSTT had not describedthe details of their procedures. The intervention

provided by rehabilitation therapists is highly expe-riential. Many training variables can be manipulat-ed. The strategy, however, had to be given subjectsin a reproducible fashion for SCILT to succeed. Forthe RCT to have internal consistency within andacross CUs and provide generalizable outcomes,the investigators developed a set of definedapproaches for BWSTT. The CU investigators met

at UCLA with additional experts involved in devel-oping BWSTT, including Drs Anton Wernig (Berlin)and V. Reggie Edgerton (UCLA). The literaturerevealed that treadmill speeds had been less thanone-third of normal casual walking velocity in priorstudies, and weight support was rapidly eliminat-ed.7 Such slow training speeds presumably did notprovide typical locomotor-related sensory inputs. Aconsensus was reached to aim for training at tread-mill speeds as close as feasible to near normalcasual walking speeds, approximately 2 to 2.5mph, and to continue some level of body weightsupport to help achieve these speeds. As soon asfeasible, the level of weight bearing was increasedin increments within and between training ses-sions. A complimentary over-ground walking pro-gram was also established. The Trainers Group cre-ated and revised a training manual and provided a1-week course for all participants at UCLA.34 Atleast 1 therapist from each CU was trained, fol-lowed by on-site training of at least 2 therapists and2 therapy assistants.A certification process followed. The BWSTT

team at each site completed a practice schedulewith at least 4 volunteer subjects with SCI over 3

months, monitored by video each week by theTrainers Group. The video was assessed for howthe therapists assisted leg, hip, and trunk control;hand placement on the legs; kinematics; and theuse of varying treadmill speeds and levels of

weight support. The trainees self-critiqued theirefforts by checklist and described their decisionmaking during the sessions. The therapists wereencouraged to discuss training techniques amongthemselves via e-mail and the study website. If theTrainers Group, using a video skills check list,found that the therapists scored above 80% in theirtechnique and problem-solving skills, a member of

the Trainers Group visited the site, did some addi-tional training over 2 days with at least 4 differentsubjects with SCI, and certified the physical thera-pists and their assistants at the site. If the therapistsscored between 60% and 80%, feedback was givenand practice continued, followed by a resubmis-sion of the videotape. If a CU did not reach the80% level, a member of the Trainers Group visitedthe site. In a trial that would last 2 years, turnover

B. H. Dobkin et al.




8/16

of therapists was expected. Training certificationwas required when a new primary therapist joineda site.A number of commercial systems for treadmill

training were available. The UCLA group workedwith the Vigor Company (Stevensville, MI) to mod-ify its body weight support system with a gas cylin-der so it offered vertical displacement appropriate-

ly timed to the step cycle and helped controlground reaction forces during stance. The treadmill

was raised on a platform and fitted with mobileseats so that therapists could sit with some comfortand flexibility as they manually assisted the sub-jects legs. Therapists had previously reportedoccasional hand and back injuries from their repet-itive lifting and reaching activities during BWSTT.

The PI and SCU visited each site to review train-ing accomplishments, data collection and entrymechanisms, scheduling of outcome measures, useof the study Web page, and financial issues. Thecoordinator at each site discussed subject identifi-cation procedures and the process of informedconsent. Information from each site was sharedamong all CUs, especially how problems wereengaged and resolved. Each site filled out dataforms and carried out each entry and follow-upprocedure for 3 to 4 subjects prior to the officialstart of the RCT.

Blinded Observer Training

Since the subject and therapists could not be

blinded regarding the intervention, a single-blinddesign was chosen. The investigators considered a

variety of ways to obtain reliable outcome data thatwould be assessed by someone who was unawareof the treatment assignment. Whereas videotapinga therapist carrying out the examination for out-come measures and having the tapes judged byone or more blinded observers could serve thispurpose, the approach would add considerablecost and complexity to the RCT. Since the outcomemeasurement tools were chosen in part for theirhigh reproducibility, the investigators decided thata trained, experienced blinded observer at each site

could obtain the measures reliably. The observerwas masked as to the subjects assignment by hav-ing all participants, including the subjects, agreenot to mention anything about the intervention andby choosing an observer who worked away fromsites of patient care.

The blinded observers were physical or occupa-tional therapists or rehabilitation nurses qualified toperform the FIM by the Uniform Data Systems

Users Group. A backup observer was also trained.If a new observer was needed over the course ofSCILT, the coordinator and lead therapist at the CUtrained that person. The PI visited each site andreviewed local procedures for testing subjects andtransferring data directly to the SCU. The PIreviewed a videotape of the testing procedures car-ried out by each blinded observer on the subjects

who were volunteers for the CU therapists as theypracticed BWSTT under the guidance of theTrainers Group. The blinded observers were grad-ed on their performance and practiced the stan-dardized data-gathering techniques until rated ascertified by the PI.

The blinded observers were the only persons tocollect outcome data relevant to the primary andsecondary hypotheses. They were notified by thecoordinating physical therapist as soon as a subject

was randomized. Entry testing was performed with-in 24 hours of randomization or on the Monday fol-lowing a Friday randomization. The observersreceived a calendar from the SCU that noted thedate that follow-up testing was expected for allsubjects at that site. The local coordinator receivedthe same calendar and arranged a convenientschedule for the subject and blinded observer. Theobserver kept open 2 mornings a week to test sub-jects.

Screening and Recruitment

Patients admitted to SCILT sites for rehabilitation

after incomplete SCI typically have to be medicallystable enough to tolerate at least 3 hours of thera-pies a day, need more than minimal assistance forambulation and self-care, are motivated to increasetheir functional independence, have adequate cog-nition to participate in treatments, and have ade-quate social supports to anticipate that they wouldreturn home on discharge. Discharge from the sitesdepends on when a patient becomes independentenough to be managed at home or when importantfunctional gains have reached a short-term plateau.

Within these community practice patterns, a proto-col for screening, recruitment, and treatment was

designed for the trial.As close to the beginning of their formal inpa-

tient rehabilitation program as medically feasible,all patients with a recently acquired traumatic SCIadmitted to 1 of the CUs was screened by the sitesphysician and coordinator to determine whetherthey qualified to participate. A form that listedinclusion and exclusion criteria for the RCT wasfilled out by the physician or coordinator for every





9/16

admission who had an incomplete traumatic SCI.Candidates who met initial criteria on admission

were introduced to the study by the physician orcoordinator. The CU personnel presented the back-ground that led to the study and read the consentform to the eligible subject. Explanations about theRCT were written in English, French, and Spanish,along with informed consent forms and the self-

completed questionnaires in the languages neededby each CU. The centers tried to make participationpractical, for example, by assisting with transporta-tion for outpatient therapy and follow-up testing. Ifa patient agreed to consider participating, theSCILT therapist would introduce the patient to theapparatus and give the patient the opportunity tostand upright attached to the harness if so desired.

The study was not marketed beyond flyers at eachsite. All local publicity was limited to a simple, bal-anced statement about the trial that was authorizedby the PI in consultation with the investigators. TheSCU maintained a Web site for the public thatdescribed the trial and eligibility criteria and listedparticipating sites. No lectures or publications bythe SCILT group were permitted to go beyond therationale and specific aims of the RCT. The PI andsite investigators decided that the RCT ought to beconducted out of the spotlight that many trials cre-ate to avoid misinterpretations by the public aboutthe potential efficacy of BWSTT and to dissociatethemselves from the marketing efforts of commer-cial entities that make equipment or offer their ownfor-profit style of therapy. Several members of theSCILT Trainers Group, however, did commercialize

a training protocol late in the study.

Randomization

Patients who met entry criteria and consented toparticipate were randomized within 24 hours, or onthe Monday following a Friday admission. Onreceipt of the initial identifying data and inclu-sion/exclusion checklist by fax, the SCU random-ized and stratified subjects by whether they had anUMN or LMN lesion and by whether they wererated on the ASIA as B, C, or D. Stratification by CU

site was not feasible; too few subjects would havebeen entered into each cell. A random, permutedblock design with those factors was used, along

with a system to handle imbalances in assignmentsto each treatment arm within each site. By fax withe-mail corroboration, the SCU sent the assignmentand a patient study number that was the only wayto identify subjects once assigned to the CU. The

receiving CUs coordinator acknowledged receiptwith a return e-mail confirmation. Each CU kept aflow chart of these interactions.

Interventions

Subjects received 12 weeks of inpatient and out-patient therapy within the study. Based on the priorexperience of the CUs, most subjects were expect-ed to enter into the trial by 28 days post-SCI with arange of 10 to 35 days. The investigators anticipat-ed that the study intervention would end, on aver-age, by 16 weeks after onset of SCI. A minimum of45 days and maximum of 60 days of therapy wererequired of subjects.

During the inpatient rehabilitation stay, each sub-ject received at least 1 hours of daily physicaltherapy 5 days a week. As an outpatient, each sub-ject received at least 1 hour of mobility-related ther-

apy 3 days a week in addition to other necessarytreatments. Within the level of their tolerance forexercise, subjects randomized to BWSTT received 1hour of their therapy time devoted to step training

with the experimental intervention. The amount oflocomotor-related physical therapy was recordedduring the intervention period. The time spent instanding and stepping activities was recorded dailyduring the inpatient stay by the treating physicaltherapist and weekly by each subject during out-patient therapy up to the 12-week assessment.Physical and occupational therapy stressed typicalactivities outside the locomotor training time, such

as self-care skills, transfers, and wheelchair mobility.Subjects randomized to conventional therapy

alone could continue at a facility closer to theirhome or with a home health agency if they wouldotherwise not be able to complete the study. Thistreatment was monitored by the CU physician.From the end of the intervention to the final eval-uation, patients in the conventional and BWSTTgroups were given a routine home locomotor ther-apy program that encouraged them to walk 3 timesa day for at least 5 minutes in subjects with theleast ability to walk. Between the end of the inter-

vention and the 1-year follow-up, some patients

continued therapy, but they were asked not to per-form BWSTT.All standing- and walking-related exercise ses-

sions for the conventionally treated subjects wereconducted by a senior physical therapist who

worked on the SCI unit. The SCILT-trained physicaltherapist provided BWSTT and over-ground walk-ing interventions.

B. H. Dobkin et al.




10/16

BWSTT arm of the study. Each session lasted up to60 minutes, which included warm-up and stretch-ing. The length of bouts of stepping on the tread-mill depended on the subjects endurance andcapability, but 1 goal was to gradually increase to20 minutes of continuous step training. A physicianassociated with the study was available to assistshould medical complications arise. Heart rate and

blood pressure were monitored. The interventionwas not aimed at having a conditioning effect.Earlier mobilization permitted by BWSTT couldhelp prevent deconditioning even at submaximalexercise levels, however. Heart rates generally didnot exceed 110 beats/min, and patients were not tofeel more than minimally short of breath. Patients

were not advanced to higher treadmill speeds orless weight support if tachycardia or dyspneaoccurred. The physician and physical therapistmonitored subjects for untoward effects such ashypotension, pain, and skin abrasions.

Subjects received verbal instruction as well asphysical assistance as needed to practice trunk,pelvic, hip, knee, and ankle control. One trainer satby each paretic leg and facilitated its movement byplacing 1 hand above the popliteal fossa and theother above the heel. Another assistant trainerstood behind with hands on the patients hips tofacilitate trunk extension, pelvic rotation, weightshifting, and hip extension. Mirrors placed in frontand along one side of the patient provided visualfeedback about trunk and leg position for subjectsand trainers. Optimal hip extension at the end ofstance was facilitated by the patient, therapist, and

passively by treadmill belt motion. Optimal footclearance and placement, along with weight bear-ing that prevented knee hyperextension or give-

way were emphasized, as well as a step-throughpattern with symmetry of the legs in step lengthand stance time. Care was taken to limit excessiveground reaction forces when the leg in swing madefoot contact. Since sensory information from the soleand the ankle helps signal transitions from stance toswing, AFOs were not employed during BWSTT.The ankle and foot were supported by the thera-pists or with a figure-of-eight wrap for foot clear-ance at toe off and heel strike to protect the foot.

The treadmill had no side railings. Reciprocalarm movements were encouraged. Trekking polesheld parallel to the ground by assistants were usedto initially aid reciprocal arm movements. Subjects

were also encouraged to train over ground withreciprocal arm movements, using the poles heldparallel to the floor by assistants if needed, ratherthan training, for example, with a walker. Bracing

with AFOs and use of assistive devices were per-mitted during training over ground if needed forsafety and to increase independence.

Training began at from 20% to 50% body weightsupport and at treadmill speeds of at least 1.6 mph.Treadmill speeds were varied within each sessionbut aimed to reach belt speeds of at least 2 to 2.5mph (68 m/min) as soon as feasible. The level of

weight support was adjusted within sessions toachieve knee extension during stance and to loadthe stance leg as much as tolerated without causingthe knee to buckle. Subjects were advanced tohigher treadmill velocities and more full-limb load-ing until the maximum speed at which each patientexhibited his or her best step pattern at full weightbearing was achieved.

Conventional arm. The treating physical therapistsat each CU were informed about the subjects par-ticipation in SCILT and instructed to offer as muchmobility-related standing and walking activity asfeasible, but a minimum of one-half hour per dayof therapy. The therapists at the CUs employed thefollowing general strategies. The subjects classifiedas ASIA B did weight bearing on a tilt table orstanding frame to meet the practice time criteria ofthe protocol. Gait interventions targeted exercisesand problem solving for balance, weight bearing,leg symmetry in swing and stance times, selectivemuscle strengthening, and improved motor control.Forward progression of the lower extremities waspracticed at the parallel bars and over ground,using facilitatory and inhibitory techniques to

encourage reciprocal stepping and assistive andorthotic devices as needed, including walkers,Lofstrand crutches, or one or two canes when thepatient needed these for safer and more independ-ent over-ground walking.

Outcome Measures

Measurement tools for the primary and second-ary outcomes were chosen for their reliability, like-ly validity in relation to mobility outcomes, admin-istrative ease, and minimal cost. Few tools, howev-

er, have been validated in trials of a physical ther-apy for walking after SCI. Formal gait analyses

were not obtained routinely because of the cost ofkinematic, kinetic, electromyographic, and tempo-ral gait pattern data, as well as difficulty in deter-mining the most meaningful data to statisticallyanalyze. The Canadian CUs obtained gait studies inabout 10 willing participants. In addition, scales





11/16

were added on an exploratory basis to test theirreliability and validity in patients with SCI. The PIdiscussed the content of various tools with 10 of hispatients with SCI who used a wheelchair for 50% to100% of their mobility needs. With this advice andinput from the CUs, the tools chosen reflectedhome and community activities and quality of lifethat could be affected by BWSTT, such as walking

compared to wheelchair use, greater independencein walking, more functional walking velocity, andimproved trunk control.

DESCRIPTORS

Initial descriptors about each subject wereobtained by the CU coordinator at the time theinformed consent was signed. Descriptor informa-tion on entry to the study included age, sex, heightand weight, residence, occupation, medications,cause of SCI, neurologic level of injury and ASIAclassification on entry to the study and at the time

of injury if available, time from onset of injury toentry into the study, and FIM locomotor score. Thetotal Lower Extremity Motor Scores were collectedby the SCILT physical therapist or physician atentry and by the blinded observer. This scaleassigns a grade for strength (0 = no movement, 5 =normal resistance) to 5 key muscle groups of eachleg, each with a different dominant root level, for atotal possible score of 50. The upper extremities

were also graded similarly for a total score of 50.The SCU reviewed these for consistency, especial-ly for differences that could alter the initial ASIAgroup assignment. The PI reviewed and arbitratedinconsistent reports.

Using the ASIA standards of neurologic classifi-cation, the CU physician recorded the sensory andmotor scores at entry, at the end of the interven-tion, and then at 6 and 12 months after entry. Oninpatient discharge, the type of residence and totalrehabilitation inpatient and, later, outpatient thera-py days were recorded.

PRIMARY OUTCOMES

The tests most relevant to walking (walking

speed and FIM level) were performed by the blind-ed observer. They were collected at entry and at 2-week intervals until the end of the intervention andthen at 6 and 12 months. Measures were taken aftera brief warm-up in the morning when the patient

was not fatigued. AFOs and assistive devices wereused if required for reciprocal stepping.

1. Subjects classified as ASIA B and C on entry wererated for level of independence for walking on a

level surface for at least 50 feet if physical assis-tance was needed or for 150 feet if no helper wasrequired, using standardized criteria from the FIM7-level locomotor subscore. For ASIA D subjects,this FIM score is a secondary outcome.

2. Subjects classified ASIA D on entry were timedperforming a 50-foot walk with a reciprocal gaitpattern over a flat, straight-tiled surface. The startand stop points were shown to the subject. The

subject started from a standing position 2 stepsbehind the start line. The instruction to begin towalk was ready and go. As the lead foot crossedthe start line, the stopwatch was started. Thepatient was encouraged to walk as fast as he orshe felt safe. The blinded observer walked along-side the subject and provided encouragement andcontact guarding as needed. Only 1 other personcould assist the subject. The observer graded thelevel of assistance given. The walk ended whenthe lead foot crossed the finish line. Two trialswere performed with a 5-minute rest betweenwalks. The heart rate was measured at the start andfinish, as a marker for the energy cost of locomo-tion. For ambulatory ASIA B and C subjects, speedis a secondary measure.

The primary outcome measures were also usedas stop points for the assigned arm of therapy. Theduration of conventional or BWSTT treatment wasa minimum of 6 weeks in subjects who reached theendpoint of an independent level of reciprocal gait(FIM locomotor score of 6 or 7) and a normal casu-al walking speed of 50 feet in 15 seconds (equiva-lent to 100 cm/sec, 60 m/min, or 2.2 mph). Thesemilestones indicate the potential for highly func-tional community ambulation, a level that relative-ly few patients ordinarily achieve during their reha-

bilitation. Training beyond 6 weeks, for up to 12weeks, was continued for subjects in either armwho had not achieved a level of independence of6 or 7 on the FIM locomotor subscale and who did

walk 50 feet in 15 seconds or less.

SECONDARY OUTCOMES

1. Endurance was assessed by measuring the distanceachieved in a continuous walk with a reciprocalgait pattern for 6 minutes. The test was performedon a flat, uncarpeted, 100-foot walkway. Thus, thewalk included turns of 180 degrees. The blindedobserver walked alongside the subject and provid-

ed encouragement and contact guarding as need-ed. Physical walking assistance was provided withthe help of 1 person. The observer graded thelevel of assistance given. A chair was placed at the50-foot midpoint of the walkway or an aide fol-lowed behind the subject with a wheelchair forsafety purposes should the subject suddenlyfatigue. Tests of walking endurance are reliableindicators and can be more sensitive to changeduring rehabilitation than the FIM walking sub-score or speed of ambulation.35 These data were

B. H. Dobkin et al.




12/16

collected at the end of the intervention (12 weeks)and at the 6- and 12-month assessments. The BergBalance Scale assesses other potential effects ofthe interventions especially on chair transfers andsitting and standing balance.36 During the 50-footwalk, the Walking Index for Spinal Cord Injury(WISCI) provided a hierarchical view of theamount of physical, bracing, and assistive devicehelp needed by subjects to walk.37 The WISCI has

face validity, concurrent validity, and interraterreliability. For the RCT, a zero level was added,splints could be substituted for braces, and thebraces were described as locked or unlocked.38

The blinded observer assessed these mobilitymeasures at entry, 4 weeks, 12 weeks, 6 months,and 12 months.

2. A variation of the Rand Medical Outcomes StudySF-36 served as the primary tool used to measureany changes in quality of life. It reflects patientperceptions and preferences about physical, emo-tional, and social well-being. The tool has beenemployed in studies of patients with chronic dis-eases. No studies at the start of the RCT had beenreported to show its validity in the population thatthe trial studied and, in particular, in how inde-pendence in ambulation relates to quality of life,but this measurement strategy has received con-siderable support.39,40 The SF-36 was modified withsome additional questions found to be useful inpatients with multiple sclerosis, called the SF-54.40

Subjects are able to fill out this form in 10 to 15minutes. Within the physical functioning portion ofthe scale, 2 items were altered to better reflect thecapacities of persons with SCI. The Reintegrationto Normal Living (RNL)41 and an Activities CheckList (ACL) of 25 home and community activitiesthat include hobbies, filled out by each subject,and the total FIM scores, completed by the blind-ed observer, also reflect aspects of quality of life,self-care, and community levels of independence.

The Beck Depression Scale, completed by sub-jects, further assessed mood. The total FIM motorscore was collected at entry, 12 weeks, and at 6and 12 months. The SF-54, RNL, ACL, and Beckwere collected at the 3, 6, and 12-month assess-ments, when the subjects are able to try to inte-grate into the community.

3. The initiation of even partially supported steppingby BWSTT could help prevent medical complica-tions of early immobility. Problems include pneu-monia, thrombophlebitis, pulmonary emboli, jointcontractures, muscle atrophy, deconditioning,dysautonomia, pressure sores, osteoporosis, het-erotopic ossification, leg and truncal spasms, anddepression. Potential complications of the BWSTT

intervention include an increase in spasticity, skinlesions, and joint or other limb pain. We used theModel Systems SCI Program criteria for definingmedical complications such as decubitus ulcers,pneumonia versus atelectasis, deep vein thrombo-sis, and pulmonary emboli. The physician investi-gator recorded adverse reactions and complica-tions as they happened. The PI and the SafetyCommittee reviewed all complications entered onthe forms completed by the sites physician andphysical therapist.

Medications used to treat spasms and rigiditymay affect neurotransmitters in the cord and, intheory, alter spinal learning. The unnecessary useof these drugs may confound locomotor-relatedoutcome measures. Subjects and their treatingphysicians were encouraged not to use anti-spas-ticity agents, unless the medications had a clear-cutbenefit on spasms that interrupted sleep or wheel-

chair activities or prevented safe positioning. Theirunavoidable use during therapy or at the time ofoutcome measures was recorded. Reliable meas-ures of hypertonicity employed included the

Ashworth Scale and the number of flexor andextensor spasms counted by each subject on theday prior to the assessment of the blinded observer.These measures of spasticity, however, do not re-flect the effects of hypertonicity on mobility-relatedactivities.

DropoutsIf a subject developed a medical complication

that prevented participation in rehabilitation mobil-ity training for more than 4 days, the length oftraining was extended for that lost time, for up to 2consecutive weeks. This information was recordedand sent to the SCU. If the physician investigator atthe site determined that continued BWSTT or con-

ventional therapy was medically contraindicatedfor more than 2 weeks, the subjects situation wasdiscussed with the PI and a decision made on

whether the subject should continue in the study.

Since the RCT planned an intention-to-treat analy-sis, every effort was made to keep subjects in theprotocol. The SCU was informed about any dis-continuance and the reasons on a standard form.

Plans for Retention and Compliance

Communication and support among sites wascritical for maintaining the stamina and diligenceneeded to complete the RCT. The investigatorsencouraged the therapists, Training Group, blindedobservers, and coordinators to share problems and

solutions. Each CU used its commitment of institu-tional support to foster an understanding of ran-domization and of the 2 arms of the RCT for itsstaff. Hospital staff was trained to understand theneed for the trial and to acknowledge that until theRCT was completed, no one should comment onperceptions of the efficacy of either arm. The thera-pists who provided conventional therapy were espe-cially encouraged to follow the training protocol.





13/16

A public website provided information about thetrial. The nonpublic website allowed CU physi-cians, therapists, coordinators, and blinded observersto ask questions and provide answers. The PI usedthe site to keep CUs aware of developments andmeetings. The SCU managed the website. To add abit of competition between sites, the website pro-

vided monthly summaries of the number of sub-

jects within the UMN and LMN subgroups at eachsite, a monthly summary of number of subjectsentered compared to the number expected by thatdate, and the number who completed their 6- and12-month assessments. The SCU also reviewed thisinformation with each CU to keep the trial onschedule for recruitments and assessments. TheSCU provided each coordinator and blindedobserver with a calendar of the anticipated date offollow-up assessment shortly after a subject hadbeen randomized. The follow-up reminder wasreinforced by e-mail, at least 2 weeks in advance.In addition, questions or answers that were not ofgeneral interest were relayed by e-mail, and impor-tant developments, such as adverse reactions, weremade known to all by e-mail. Two hundred andfifty e-mail messages to and from the PI wereexchanged over the first 3 years of the RCT.

Each site shared ways to retain its subjects. Thecoordinator visited subjects during the inpatientstay and called every 2 weeks of outpatient careand monthly before the 6- and 12-month assess-ments. Every effort was made to accommodate asubjects personal schedule for follow-up visits.Subjects who needed help to defray travel expens-

es to complete their last 4 weeks or final 12 visitsto a CU for outpatient therapy received $20 pervisit for transportation. Subjects were paid $50 toreturn for each of 3 outcomes assessments. Thispayment covered travel expenses and helpedensure compliance with these critical assessments.

Data Entry and Management

An operations manual produced at UCLA provid-ed detailed information and flow charts for all pro-cedures related to the conduct of the trial. The

manual included the methods for BWSTT andproblem-solving algorithms for its application.

The physical therapist or coordinator at each CUsent completed forms to the SCUs data entry clerkas soon as the scannable forms were completed.The information was not part of a patients inpa-tient or outpatient hospital record. The independ-ent blinded observers measures were sent directlyto UCLA. The coordinator at each of the CUs main-

tained a folder with the name and address of eachsubject, along with all baseline data collected bythe physician and coordinator.

The SCU maintained a database containing thepatient identification number and treatment assign-ment that will be merged with other study variablesat the end of the trial. A separate database was cre-ated and data were entered and saved using

Microsoft Access. These files and the separatepatient logs will be converted into SAS format forstatistical analyses. The database software includesprograms to check whether entered data fall intoplausible ranges and whether data are complete.Data entries were checked by the SCU for accura-cy by repeating the entry procedure and lookingfor discrepancies. The primary data manager con-tacted a coordinator or blinded observer at a CU ifdata on the forms appeared inconsistent, for exam-ple, if the impairment group assignment is incon-sistent with the motor examination on the ASIA or

if the FIM locomotor score is inconsistent with thetime to do the 50-foot walk.

Unblinding Procedures

The external Safety and Data MonitoringCommittee, appointed by the PI and the NIH pro-gram director, served as an independent oversightbody. To optimize its interactions, the committee

was drawn from academic sites in Los AngelesCounty but not from UCLA. The committee moni-tored the progress of patient accrual, data manage-

ment, patient safety, and scheduled and requestedinterim analyses of data. The Safety Committee wasblinded as to the therapy assignment of subjectsbut had the option of becoming unblinded ifadverse reactions threatened the safety of subjects.

Data Analyses

The statistical design included an interim analysiswhen one-half of the planned number of subjectshad a 6-month assessment for the primary out-

comes. At that stage, the Safety Committee deter-mined if the study appeared to demonstrate effica-cy or lack of efficacy for one arm compared to theother within the ASIA B, C, or D groups and over-all. They assessed the potential futility of continu-ing to enter subjects if the conditional power wastoo small to expect that additional accrual wouldresult in significant effects (that is, rejection of thenull hypothesis).

B. H. Dobkin et al.




14/16

The initial protocol emphasized the developmentof valid statistical models to carry out data analyses.SCILT is essentially 2 clinical trials: ASIA B and Ccombined and ASIA D since the primary endpointis different for each ASIA stratificationwalkingindependence score (FIM walking scale) and walk-ing speed. The statistical design for the first trial isas follows:

Asia Lesion BWSTT Conventional

B UMNB LMNC UMNC LMN

This design has 2 stratification factors and 1 treat-ment factor with repeated measurements at base-line and biweekly for the first 12 weeks, then at 6and 12 months. The primary endpoint for trial 1 isthe FIM walking score at month 6. The primary

analyses consist of an ANCOVA model for the FIMwalking score at month 6 as follows:

Yijkl6

= i+_

j+_

k+

l+ (_)

jl+

(_)kl

+ *Yijkl0

+ *Xijkl

+ _ijkl

i= 1, 2, . . . n

jkl individual

j= 1, 2 ASIA B versus C

k= 1, 2 UMN versus LMN

l= 1, 2 BWSTT + conventional versus conventionalY

ijkl0= FIM walking score at baseline (covariate)

Xijkl

= duration from onset ofSCI to study entry (covariate)

_ijkl

= error term for the individual observation

The resulting ANCOVA will enable us to studypossible treatment by stratification interaction,main effects, and initial covariates. To carry out thisanalysis, detailed diagnostics for model adequacy,goodness of fit, outlier observations, and so forth

will be carried out. A similar model will be devel-oped for the second clinical trial for patients classi-fied as ASIA D.Analyses for the secondary endpoints will use the

appropriate models. Correlation among the sec-ondary and primary variables will be obtained foreach clinical trial. For example, over the course ofthe RCT, we will examine relationships between

the primary outcomes of locomotor independenceand walking speed with walking endurance, themaximum heart rate after walking 50 feet, level ofinjury, changes in ASIA classification, the LowerExtremity Motor Scores, and the WISCI. We willlook for aspects of quality of life that are reflectedin locomotor capacity. This will be considered inseveral ways, including a bivariate ANCOVA. The

full database includes measurements of endpointsmade over time. A repeated measures ANCOVAusing alternative nonparametric time trend curveestimation methods will be carried out making useof generalized estimating equation methodology.

DISCUSSION

Multicenter RCTs of physical therapies have beenan uncommon undertaking for those who practiceneurologic rehabilitation. This problem derivesfrom confounding issues. Few neurophysiological-ly sound hypotheses have worked their way intothe clinical arena. Few styles of physical therapyare carried out in a reproducible fashion acrossrehabilitation centers. Pharmacologic interventions,especially for chronic symptoms or signs such asspasticity, outdistance studies of physical therapies,in part because dose-response curves are deter-

mined by preliminary drug studies, the interventionis precise, and outcome measures are kept simple,perhaps too simple to reveal much clinicalimpact.42 Outcome measures pose their own limi-tations. Many tools that examine disability, impair-ment, participation, and quality of life have floorand ceiling limitations or are too broad for discern-ing the benefit of an intervention for a well-definedproblem.1

The SCILT investigators believe they have stan-dardized a conventional and an experimental phys-ical intervention as best as reasonably possibleacross sites and therapists. Differences in outcomes

between sites, of course, would degrade this confi-dence. Parametric and nonparametric outcomemeasures used in the RCT ought to provide clini-cally meaningful changes if, indeed, BWSTT is effi-cacious at an early stage after SCI at the intensityand duration applied. The results will also be rele-

vant to people with chronic or complete SCI who,in the future, may receive biological interventionsthat will have to be combined with locomotor train-ing to enhance gains for walking.28

Recent trials of well-defined interventions forneurologic rehabilitation, such as this RCT and theupper-extremity practice intervention for narrowly

defined patients with stroke called EXCITE, openthe door for planning other multicenter trials,regardless of the efficacy of either of these pio-neering efforts. These trials demonstrate thatprospective, randomized multicenter trials withmasked outcomes can be organized to test a phys-ical intervention in a representative population andcan be managed to completion.





15/16

ACKNOWLEDGMENTS

We thank Drs Beth Ansel and Michael Weinrichat the NIH/NICHD for their input into the grant(HD 37349) funded by their agency, the NationalCenter for Medical Rehabilitation Research. Biodex(Shirley, NY) rehabilitation treadmills and RobertsonHarness (Henderson, NV) adjustable medical har-

nesses were used.

REFERENCES

1. Dobkin B. The clinical science of neurologic rehabilitation.New York: Oxford University Press; 2003.

2. Marino R, Barros T, Bioering-Sorensen F, et al. Internationalstandards for neurological classification of spinal cordinjury. 6th ed.J Spinal Cord Med2003;26(suppl 1):S49-56.

3. Stover S, DeVivo M, Go B. History, implementation, andcur rent status of the national spinal cord injury database.Arch Phys Med Rehabil1999;80:1365-71.

4. Dobkin B, Apple D, Barbeau H, Saulino M, Fugate L, ScottM. Randomized trial of body weight-supported treadmill

training after acute spinal cord injury.Neurorehabil NeuralRepair1999;13:50.

5. Barbeau H, Fung J. New experimental approaches in thetreatment of spastic gait disorders. Med Sport Sci1992;36:234-46.

6. Barbeau H, McCrea D, ODonovan M, Rossignol S, Grill W,Lemay M. Tapping into spinal circuits to restore motorfunction.Brain Res Rev1999;30:27-51.

7. Dobkin B. Overview of treadmill locomotor training withpartial body weight support: a neurophysiologically soundapproach whose time has come for randomized clinical tri-als.Neurorehabil Neural Repair1999;13:157-65.

8. Wernig A, Nanassy A, Miller S. Maintenance of locomotorabilities following Laufband (treadmill) therapy in para-and tetraplegic persons: follow-up studies. Spinal Cord1998;36:744-9.

9. Dietz V, Wirz M, Curt A, Colombo G. Locomotor patterns

in paraplegic patients: training effects and recovery ofspinal cord function. Spinal Cord1998;36:380-90.

10. Tillakaratne N, de Leon R, Hoang T, Roy R, Edgerton V,Tobin A. Use-dependent modulation of inhibitory capacityin the feline lumbar spinal cord. J Neurosci 2002;22:3130-43.

11. Schmidt B, Jordan L. The role of serotonin in reflex mod-ulation and locomotor rhythm production in the mam-malian spinal cord. Brain Res Bull2001;53:689-710.

12. Kakulas B. A review of the neuropathology of humanspinal cord injury with emphasis on special features. JSpinal Cord Med1999;22:119-24.

13. Blight A. Remyelination, revascularization, and recovery offunction in experimental spinal cord injury. In: Seil F, edi-tor. Spinal cord injury. New York: Raven; 1993, p. 91-104.

14. Belanger M, Drew T, Provencher J, Rossignol S. A compar-ison of treadmill locomotion in adult cats before and after

spinal transection.J Neurophysiol1996;76:471-91.15. Bouyer L, Whelan P, Pearson K, Rossignol S. Adaptive

locomotor plasticity in chronic spinal cats after ankleextensors neurectomy.J Neurosci2001;21:3531-41.

16. de Leon R, Hodgson J, Roy R, Edgerton V. Locomotorcapacity attributable to step training versus spontaneousrecovery after spinalization in adult cats. J Neurophysiol1998;79:1329-40.

17. Dimitrijevic M, Gerasimenko Y, Pinter M. Evidence for aspinal central pattern generator in humans. Ann NY AcadSci1998;860:360-76.

18. Duysens J, Van de Crommert H. Neural control of locomo-tion: The central pattern generator from cats to humans.Gait and Posture1998;7:131-41.

19. Tresch M, Saltiel P, Bizzi E. The construction of movementby the spinal cord. Nature Neurosci 1999;2:162-7.

20. Harkema S, Hurley S, Patel U, Dobkin B, Edgerton V.Human lumbosacral spinal cord interprets loading duringstepping.J Neurophysiol1997;77:797-811.

21. Armstrong D. The supraspinal control of mammalian loco-motion.J Physiol1988;405:1-37.

22. Dobkin B, Harkema S, Requejo P, Edgerton V. Modulationof locomotor-like EMG activity in subjects with completeand incomplete chronic spinal cord injury. J NeurologicRehabil1995;9:183-90.

23. Dobkin B. Spinal and supraspinal plasticity after incom-plete spinal cord injury: correlations between functionalmagnetic resonance imaging and engaged locomotor net-

works. In: Seil F, editor. Progress in brain research. Vol.128. Amsterdam: Elsevier; 2000. p. 99-111.

24. Graziano M, Taylor C, Moore T. Complex movementsevoked by microstimulation of precentral cortex. Neuron2002;34:841-51.

25. Merzenich M, Recanzone G, Jenkins W, Nudo R. How thebrain functionally rewires itself. In: Arbib M, Robinson J,editors. Natural and artificial parallel computations.Cambridge (MA): MIT Press; 1990, p. 170-98.

26. Schwab M. Repairing the injured spinal cord. Science2002;295:1029-31.

27. Barbeau H, Fung J. The role of rehabilitation in the recov-ery of walking in the neurological population. Curr OpinNeurol2001;14:735-40.

28. Dobkin B. Functional rewiring of brain and spinal cordafter injury: the three Rs of neural repair and neurologicalrehabilitation. Curr Opin Neurol2000;13:655-9.

29. Edgerton V, de Leon R, Harkema S, et al. Retraining theinjured spinal cord. J Physiol2001;533:15-22.

30. Edgerton V, Harkema S, Dobkin B. Retraining the humanspinal cord. In: Lin V, editor. Spinal cord medicine: princi-ples and practice. New York: Demos Medical; 2003. p. 817-28.

31. Waters R, Adkins R, Yakura J, Vogil D. Prediction of ambu-latory performance based on motor scores derived fromstandards of the American Spinal Injury Association. ArchPhys Med Rehabil1994;75:756-60.

32. Waters R, Yakura J, Adkins R, Barnes G. Determinants ofgait performance following spinal cord injury. Arch PhysMed Rehabil1989;70:811-8.

33. Fiedler R, Granger C, Russell C. UDSMR: follow-up data onpatients discharged in 1994-1996. Am J Phys Med Rehabil2000;79:184-92.

34. Behrman A, Harkema S. Locomotion training after humanspinal cord injury: a series of case studies. Phys Ther2000;80:688-700.

35. Pohl P, Duncan P, Perera S, et al. Influence of stroke-relat-ed impairments on performance in 6-minute walk test. JRehabil Res Develop 2002;39:439-44.

36. Berg K, Maki B, Williams J, Holl iday P, Wood-Dauphinee S.Clinical laboratory measures of postural balance in an eld-erly population.Arch Phys Med Rehabil1992;73:1073-80.

37. Ditunno J, Ditunno P, Graziani V, et al. Walking index forspinal cord injury (WISCI): an international multicenter

validity and reliability study. Spinal Cord2000;38:234-43.38. Ditunno P, Ditunno J. Walking index for spinal cord injuryscale revision (WISCI II). Spinal Cord2001;39:654-6.

39. Ware J. Measuring patients views: the optimum outcomemeasure.BMJ1993;306:1429-30.

40. Vickrey B, Hays R, Genovese B, Myers L, Ellison G.Comparison of a generic to disease-targeted health-relatedquality-of-life measures for multiple sclerosis. J ClinEpidemiol1997;50:557-69.

41. Korner-Bitensky N, Wood-Dauphinee S, Siemiatycki J,Shapiro S, Becker R. Health-related information postdis-

B. H. Dobkin et al.




16/16

charge: telephone versus face-to-face interviewing. ArchPhys Med Rehabil1994;75:1287-96.

42. Brashear A, Gordon M, Elovic E, Kassicieh D, Group BS.Intramuscular injection of botulinum toxin for the treat-ment of wrist and finger spasticity after a stroke.New EnglJ Med2002;347:395-400.

THE SCILT GROUP

University of California Los Angeles, Departments ofNeurology and of Biomathematics

Principal investigator: Bruce H. Dobkin, MDStatistical Coordinating Unit (SCU): Robert Elashoff,

PhD (co-investigator), Joanie Chung, Rebecca LiuTrainers group leader: Susan Harkema, PhD (co-

investigator)

Trainers Group

Michele Basso, PT, EdD, Andrea Behrman, PT, PhD(co-investigator) (University of Florida); LisaFugate, MD, Susan Ono, PT (UCLA)

Safety and DataMonitoring Committee

Carolee Winstein, PT, PhD, and Anny Xiang, PhD(University of Southern California), PatriciaNance, MD (VA Long Beach Healthcare System),

and Beth Ansel, PhD (National Institute of ChildHealth and Development)

Clinical Unit Sites

Magee Rehabilitation Center/Jefferson University:Michael Saulino, MD (co-investigator), JohnDitunno, MD (co-investigator), Department ofRehabilitation Medicine; Amy Bratta. MPT, Mary

Schmidt-Read, PT, MSMcGill University/Institut de Readaptation de

Montreal: Hugues Barbeau, PhD (co-investigator),Department of Physical Therapy; Brigitte Bazinet,MD, Christiane Garneau, PT, Michael Danakas, PT

The Ohio State University: Lisa Fugate, MD (co-investigator), Department of RehabilitationMedicine; Michele Basso, PT, EdD (co-investiga-tor); Rachel Botkin, PT

Ottawa University Rehabilitation Hospital: DanDeForge, MD (co-investigator), Department ofRehabilitation Medicine; Jennifer Nymark, PT,Michelle Badour, PT

Rancho Los Amigos Rehabilitation Center: MichaelScott, MD (co-investigator), Department of

Rehabilitation Medicine, Jeanine Yip-Menck, PT,Claire Beekman, PTShepherd Rehabilitation Center: David Apple, MD

(co-investigator), Department of RehabilitationMedicine; Gary Dudley, PhD (co-investigator,University of Georgia); L. VanHiel, PT, ScottBickel, PT, PhD



Methods Randomized Trial Wst

Documents

Transcript of Methods Randomized Trial Wst