PrognosisforGrossMotorFunctionMGCs

Embed Size (px)

Citation preview

  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    1/8

    current as of February 22, 2009.Online article and related content

    http://jama.ama-assn.org/cgi/content/full/288/11/1357

    . 2002;288(11):1357-1363 (doi:10.1001/jama.288.11.1357)JAMA

    Peter L. Rosenbaum; Stephen D. Walter; Steven E. Hanna; et al.

    Creation of Motor Development CurvesPrognosis for Gross Motor Function in Cerebral Palsy:

    Correction Contact me if this article is corrected.

    CitationsContact me when this article is cited.This article has been cited 112 times.

    Topic collectionsContact me when new articles are published in these topic areas.

    Prognosis/ Outcomes

    the same issueRelated Articles published in

    . 2002;288(11):1399.JAMAStephen L. Kinsman.Predicting Gross Motor Function in Cerebral Palsy

    http://pubs.ama-assn.org/misc/[email protected] http://jama.com/subscribe

    Subscribe

    [email protected]/E-prints http://jamaarchives.com/alerts

    Email Alerts

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/cgi/content/full/288/11/1357http://jama.ama-assn.org/cgi/content/full/288/11/1357http://jama.ama-assn.org/cgi/content/full/288/11/1357http://jama.ama-assn.org/cgi/alerts/ctalert?alertType=correction&addAlert=correction&saveAlert=no&correction_criteria_value=288/11/1357http://jama.ama-assn.org/cgi/alerts/ctalert?alertType=correction&addAlert=correction&saveAlert=no&correction_criteria_value=288/11/1357http://jama.ama-assn.org/cgi/alerts/ctalert?alertType=citedby&addAlert=cited_by&saveAlert=no&cited_by_criteria_resid=jama;288/11/1357http://jama.ama-assn.org/cgi/alerts/ctalert?alertType=citedby&addAlert=cited_by&saveAlert=no&cited_by_criteria_resid=jama;288/11/1357http://jama.ama-assn.org/cgi/external_ref?access_num=jama%3B288%2F11%2F1357&link_type=ISI_Citinghttp://jama.ama-assn.org/cgi/alerts/ctalert?alertType=citedby&addAlert=cited_by&saveAlert=no&cited_by_criteria_resid=jama;288/11/1357http://jama.ama-assn.org/cgi/alerts/collalerthttp://jama.ama-assn.org/cgi/alerts/collalerthttp://jama.ama-assn.org/cgi/alerts/collalerthttp://jama.ama-assn.org/cgi/content/short/288/11/1399http://jama.ama-assn.org/cgi/content/short/288/11/1399http://jama.ama-assn.org/cgi/content/short/288/11/1399http://jama.ama-assn.org/cgi/content/short/288/11/1399http://jama.ama-assn.org/cgi/content/short/288/11/1399http://pubs.ama-assn.org/misc/permissions.dtlhttp://pubs.ama-assn.org/misc/permissions.dtlhttp://pubs.ama-assn.org/misc/permissions.dtlhttp://jama.com/subscribehttp://jama.com/subscribehttp://jama.com/subscribemailto:[email protected]:[email protected]:[email protected]://jamaarchives.com/alertshttp://jamaarchives.com/alertshttp://jamaarchives.com/alertshttp://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/mailto:[email protected]://jamaarchives.com/alertshttp://pubs.ama-assn.org/misc/permissions.dtlhttp://jama.com/subscribehttp://jama.ama-assn.org/cgi/content/short/288/11/1399http://jama.ama-assn.org/cgi/alerts/collalerthttp://jama.ama-assn.org/cgi/alerts/ctalert?alertType=citedby&addAlert=cited_by&saveAlert=no&cited_by_criteria_resid=jama;288/11/1357http://jama.ama-assn.org/cgi/external_ref?access_num=jama%3B288%2F11%2F1357&link_type=ISI_Citinghttp://jama.ama-assn.org/cgi/alerts/ctalert?alertType=correction&addAlert=correction&saveAlert=no&correction_criteria_value=288/11/1357http://jama.ama-assn.org/cgi/content/full/288/11/1357
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    2/8

    ORIGINAL CONTRIBUTION

    Prognosis for Gross Motor Functionin Cerebral PalsyCreation of Motor Development CurvesPeter L. Rosenbaum, MD, FRCPC

    Stephen D. Walter, PhD

    Steven E. Hanna, PhD

    Robert J. Palisano, ScD

    Dianne J. Russell, MSc

    Parminder Raina, PhD

    Ellen Wood, MD, FRCPC, MSc

    Doreen J. Bartlett, PhD

    Barbara E. Galuppi, BA

    CEREBRAL PALSY OCCURS IN EV-ery to 2/1000 to 2.5/1000 livebirths.1 It is . . . an umbrellaterm covering a group of non-

    progressive, but often changing, motorimpairmentsyndromes secondary to le-

    sions or anomalies of the brain arisingin the early stages of development.2

    Thus, whatever additional developmen-tal difficulties individuals with cerebralpalsy might have as a result of impair-ment of the developing central nervoussystem, the hallmark of these condi-tions is a disorder in the development ofgross motor function.

    When first told that their childhas ce-rebral palsy (generally in the childs first18 months of life),parents usually wantto know its severity and whether their

    child will ever be able to walk. The evi-dence on whichto baseanswers was,un-tilrecently, limitedto observationsaboutthe associationbetween constellations ofreflexand earlymotor skillsatage2 yearsand walking at a later age3; or on motor

    milestones such as sitting between theages of 2 and 4 years and walking at alater age. However, thefindings based oneven these simple markers are conflict-ing.4,5 Crude estimates of the probabil-

    ity of being able to walk 10 steps un-

    aided at or after age 5 years vary fordifferent clinicaltypes of cerebralpalsy.6

    These observations derive from clinicsamples and are likely not representa-tive of the entire population of children

    with cerebral palsy.

    Author Affiliations: Departments of Pediatrics (DrRosenbaum) and Clinical Epidemiology and Biostatis-tics(Drs Rosenbaum, Walter, Hanna, andRaina),Schoolof Rehabilitation Science (Drs Rosenbaum and Hanna,and Ms Russell), CanChildCentre for Childhood Dis-abilityResearch (DrsRosenbaum, Walter, Hanna, Pali-sano, Raina, Wood, Bartlett, and Mss Russell andGaluppi), McMaster University, Hamilton, Ontario;Departmentof RehabilitationSciences, MCPHahnemann

    University, Philadelphia, Pa (Dr Palisano); Departmentof Pediatrics, Dalhousie University, Halifax, Nova Sco-tia(Dr Wood); andSchool of Physical Therapy, Univer-sityof WesternOntario, London, Ontario(Dr Bartlett).CorrespondingAuthor and Reprints: Peter L. Rosen-baum, MD, FRCPC, McMaster University, CanChildCentre forChildhood Disability Research, IAHS, Room408, 1400 Main St W, Hamilton, Ontario, Canada,L8S 1C7 (e-mail: [email protected]).

    Context Lack of a valid classification of severity of cerebral palsy and the absence oflongitudinal data on which to base an opinion have made it difficult to consider prog-nostic issues accurately.

    Objective To describe patterns of gross motor development of children with cere-bral palsy by severity, using longitudinal observations, as a basis for prognostic coun-

    seling with parents and for planning clinical management.

    Design Longitudinal cohort study of children with cerebral palsy, stratified by ageand severity of motor function and observed serially for up to 4 years during the pe-riod from 1996 to 2001.

    Setting Nineteen publicly funded regional childrens ambulatory rehabilitation pro-grams in Ontario.

    Participants A total of 657 children aged 1 to 13 years at study onset, representingthe full spectrum of clinical severity of motor impairment in children with cerebral palsy.

    Main Outcome Measures Severity of cerebral palsy, classified with the 5-level GrossMotor Function Classification System; function, formally assessed with the Gross Mo-tor Function Measure (GMFM).

    Results Based on a total of 2632 GMFM assessments, 5 distinct motor develop-ment curves were created; these describe important and significant differences in therates and limits of gross motor development among children with cerebral palsy byseverity. There is substantial within-stratum variation in gross motor development.

    Conclusions Evidence-based prognostication about gross motor progress in chil-dren with cerebral palsy is now possible, providing parents and clinicians with a meansto plan interventions and to judge progress over time. Further work is needed to de-scribe motor function of adolescents with cerebral palsy.

    JAMA. 2002;288:1357-1363 www.jama.com

    For editorial comment see p 1399.

    2002 American Medical Association. All rights reserved. (Reprinted) JAMA, September 18, 2002Vol 288, No. 11 1357

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    3/8

    Cross-sectional studies of motor be-havior in children with cerebral palsyhave demonstrated characteristic pat-terns of motor development accordingto severity of the condition,7 althoughthe descriptions of severity previously

    used have been crude and unsystem-atic. The motor growth curves createdby Palisano et al,8 which are based oncross-sectional population data strati-fied by severity using thevalidatedGrossMotor Function Classification System(GMFCS) for cerebral palsy,9 are an im-portant improvement.

    This articledescribespatterns of grossmotor development of a community-based sample of children with cerebralpalsy followedup prospectively. Weusedthe GMFCS to longitudinally create

    curves charting the rates and limits ofmotor function by severity of motor im-pairment. These curves increase theprognostic information available to fami-lies and clinicians considerably.

    METHODS

    Setting

    This study was made possible througha partnership between the CanChildCentre for Childhood Disability Re-search at McMaster University and the19 publicly funded regional ambula-

    tory childrens rehabilitation pro-grams in Ontario. These programs pro-vide a range of developmental therapiesand services (predominantly physical,occupational,speech-language, and rec-reational therapies) by professionalstrained and experienced in assess-ment and management of childhooddisability. Because the centers are pub-

    licly funded, each program serves themajority of eligible children in its area.

    Sample

    The sampling frame was created in early1996with18ofthe19centersand1hos-

    pital-based therapy program in a com-munity without a regional center. Eachcenter identifiedall thechildrenwho hadbeendiagnosed as having cerebral palsyand who had been born in or after 1986.Childrenwithneuromotor findings con-sistent with cerebral palsy,such as spas-ticity or reflex abnormalities, who hadnot been diagnosed as having cerebralpalsy were included in the study. Chil-dren with other neuromotor disabili-ties, such as spina bifida or muscle dis-eases, wereexcluded.Childrenwerealso

    excluded if theyhad selective dorsal rhi-zotomy,10 had received botulinum toxininjections in thelower limbs for spastic-ity management,11,12 or were receivingin-trathecalbaclofen.13 Atthe timethestudystarted, it was not yet known how theserelatively new interventionsmightaffectgross motor function. None of these in-terventions was readily available in On-tario at the timeof the study. Tothe bestof our knowledge, no children were re-ceivinghyperbaricoxygentherapy, an in-tervention that has since been shown to

    be of no added value for children withcerebral palsy.14

    Sample size calculations were per-formed using data from Scrutton andRosenbaum.7 Based on the Gross Mo-tor Function Measure-88 (GMFM-88)and estimated score limits for a 10-year-old in each GMFCS stratum (98-100, 90-95, 60-80, 12-50 and 10), a

    sample of 150 children per GMFCSstratum would provide a power of 0.85.

    Ofthesampling framecontaining2108children,1304 were stratifiedby ageandGMFCS level and were randomly se-lected.Our target was 15childrenin each

    combination of birth year and severitylevel. An initial sample was drawn fromchildren with known severity level. Wealso drew a second random sample fromthose children whose severity level wasinitially unknown. Based on the re-quired quota for each stratum, the cen-ters established the severity level for aspecified set of children. Once the levelof a child in this set became known, he/she was added to the study sample forthe appropriate stratum (TABLE 1).

    Each of the sample sizes was calcu-

    lated to achieve equal sampling frac-tions for children with initially knownor unknown severity. We oversampledfor each age and GMFCS stratum to tryto achieve at least 15 children per pre-defined cell. A total of366 children wereineligible or unavailable for various rea-sons. Of theremaining938 children,721(77%) families consented and 682(94.5%) provided data; 657 had fullyuseable data, after excluding childrenwithout cerebral palsy (FIGURE 1). Thechildren rangedinage from 1 to 13years

    at study entry.At the first assessment, therapistsreported the distribution of the childscerebral palsy as it was reported in thechildsclinicchart (hemisyndrome,diple-gia, triplegia,or quadriplegia). Theyalsoincluded any terms that had been usedto describe the diagnosis. When no for-mal diagnosishad been givento thechild,

    Table 1. Gross Motor Function Classification System (GMFCS)

    GMFCS Level

    I(n = 183)

    II(n = 80)

    III(n = 122)

    IV(n = 137)

    V(n = 135)

    Total(N = 657)

    Age, y1-2 16 13 13 12 14 68

    3-4 47 20 22 30 22 1415-6 30 15 30 30 36 141

    7-8 36 14 27 29 32 138

    9-10 36 18 20 31 26 131

    10 18 0 10 5 5 38

    Age, mean (SD) [median], y 6.90 (2.91) [6.82] 6.16 (2.75) [6.39] 6.88 (2.91) [6.85] 6.81 (2.71) [6.71] 6.76 (2.65) [6.62] 6.76 (2.80) [6.62]

    PROGNOSIS FOR GROSS MOTOR FUNCTION IN CEREBRAL PALSY

    1358 JAMA, September 18, 2002Vol 288, No. 11 (Reprinted) 2002 American Medical Association. All rights reserved.

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    4/8

    therapists were askedwhether thechildsmotor behavior and patterns lookedlike cerebralpalsy.Males(n=369)com-prised56% of thegroup.Topographicaldistribution of cerebral palsy included217 (33.0%) leg-dominant children, 62

    (9.4%) 3 limb-dominant, 263(40.0%)4

    limb-dominant, 98 (15.3 %) hemisyn-dromes, and 17 (2.8%) unknown.

    Outcome Measures

    Severity ofcerebralpalsy wasbasedsolelyon GMFCS level, whichis a reliable and

    valid system thatclassifies children withcerebral palsy by their age-specificgrossmotor activity.8,9,15 The GMFCS de-scribes the major functional character-istics of children with cerebral palsy ineach level withinthe following age win-dows: prior to second birthday; be-tween age 2 years and fourth birthday;between age 4 years and sixth birthday;and between ages 6 and 12 years. TheBOX outlines the main abilities of chil-dren aged 6 to 12years in each GMFCSlevel. Useof the GMFCS requires famil-iarity with the child, but is not a test and

    requires no formal training.Motor function was assessed with the

    GMFM.16 The GMFM is a widely used,criterion-referenced, clinical observa-tion tool with a scalefrom 0-100that wasdevelopedand validatedforchildren withcerebral palsy or Down syndrome.17 Ithas excellent reliability and demon-strated ability to evaluate meaningfulchange in gross motor function in chil-dren diagnosed as having cerebralpalsy.16,18,19 The GMFM was not de-signed to compare the function of chil-

    dren with cerebral palsy to typically de-veloping children. It measures grossmotor function in lying and rolling,crawling and kneeling, sitting, stand-ing, and walk-run-jump activities. It canbe used with any child or adolescent di-agnosed as having cerebral palsy. It fo-cuses on the extent of achievement of avariety of gross motor activities (mainlymobility skills and activities requiringpostural control such as sitting, kneel-ing, and standing on 1 foot) that a typi-cally developing 5-year-old could ac-

    complish. For data analyses, we usedscores derived from the GMFM-66, ameasurewith interval levelsthat wasde-veloped by Rasch analysis of the origi-nal 88-item scale (GMFM-88).18-20

    Procedures and Quality Control

    The ethics review boards of HamiltonHealth Sciences Corp, the Bloorview

    MacMillan Centre (Toronto, Ontario),and the Thames Valley Childrens Cen-tre (London, Ontario) approved thestudy. It was centrally managed at Can-Child, with a site coordinator in eachcenter, who was responsible for theday-

    to-day management of data collection.Before beginning to assess children, alltherapists were trained on the adminis-tration and scoring of the GMFM.21

    Their reliability was assessed against acriterion tape at the end of training andreassessed annually over the 4 years ofdata collection to ensure that they con-tinued to score the measure reliably.

    To track individual grossmotor devel-opment, children younger than 6 yearswere assessed with the GMFM-66approximately every6 months,andolderchildren were assessed every 9 to 12

    months. This timing was based on pre-vious observations that led us to expectmore rapid change in gross motor devel-opmentin thepreschool years.8 Oneachoccasion, therapists were also asked toclassify thechilds currentGMFCS level.

    Analysis

    To estimate the parameters of a non-linear model of motor development,nonlinear mixed-effects modelling22 wasused for children in each of the 5GMFCS levels. Importantly, in addi-

    tion to describing the average patternof development in each level, this analy-sis allows for orderly variations in thepatterns of development. The degree ofindividual variationswas estimated andindividual motor development curveswere fitted for each child. The modelhas 2 parametersthe estimated rateand limit of motor developmentthat have straightforward clinical in-terpretations. The model assumes thatchildren have GMFM-66 scores nearzeroat birth. Subsequently, childrenare

    expected to acquire gross motor abili-ties rapidly, with the rate of develop-ment slowing as they approachthe limitof their potential.7,8 Based on clinicalex-perience,the rate andlimit of motor de-velopment are expected to vary sub-stantially. Initial inspection of the datasuggested thatthismodel might fit thesechildren well.

    Figure 1. Sample Selection and Recruitment

    2108 Children in Sampling Frame

    1304 Randomly Selected

    938 Invited to Participate

    721 Consented to Participate

    682 Provided Data

    657 Included in Analyses

    366 Ineligible or Unavailableto Participate

    217 Refused Consent

    25 Excluded24 Ineligible

    1 Incomplete (Poor) Data

    Box. Gross Motor Function

    Classification System Levels

    for Children With Cerebral

    Palsy Between the Ages

    of 6 and 12 Years9

    Level IWalks without restrictions; limita-tions in more advanced gross motorskills

    Level II

    Walks withoutassistive devices;limi-tations in walking outdoors and inthe community

    Level III

    Walks with assistive mobility de-vices; limitations in walking out-doors and in the community

    Level IV

    Self-mobility with limitations; chil-dren are transported or use powermobility outdoors and in the com-munity

    Level V

    Self-mobility is severely limited evenwith the use of assistive technology

    PROGNOSIS FOR GROSS MOTOR FUNCTION IN CEREBRAL PALSY

    2002 American Medical Association. All rights reserved. (Reprinted) JAMA, September 18, 2002Vol 288, No. 11 1359

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    5/8

    RESULTSOver the course of the study, 657 chil-dren had a total of 2632 GMFM assess-ments, or an average of 4 observationsperchild.From these data, 5 distinct andsignificantly different motor growth

    curves, which described patterns of grossmotor development by GMFCS level,were created (FIGURE 2 and FIGURE 3).Parameter estimates for the averageGMFM-66 curve in each level are re-ported in TABLE 2. As expected, theestimated limit of development de-creased as severity of impairment in-creased. Confidenceintervals (95% CIs)for the limit parameters are tight andconfirm that each level of severity is sig-nificantlydifferent fromthe adjacentlev-els. For clinical purposes, the esti-mated variances in limit for each level

    have been used to construct intervalsthat are expected to encompass 50% ofthe limits in the population. These in-dividual differences in limit are plottedin Figure 2. The 95% CIs are conceptu-ally different from and unrelated to the

    50% bands. The 95% CIs provide an es-timate of the precision of the point es-timates ofthemean limits,whilethe50%bands provideclinicalinformation aboutthe degree to which individuals are ex-pected to vary around that mean.

    To enhanceinterpretation, theratepa-rameters from the nonlinear growthmodels have been transformed to age-90, the age in years by which childrenare expected to reach 90% of their mo-tor development potential. Smaller val-ues(inyears)indicate faster progress to-ward motor development limits.Age-90

    datainTable 2 suggest a trendfor a fasterprogression to the limit as severity of im-pairment increases. However, the 95%CIs indicate that children in levels IIIthrough V progress significantly fasterthan children in level I, but children in

    level II do not progress faster than chil-dren in level I. An earlier (younger)age-90 does not indicate better devel-opmental progressonly that a child iscloser to his/her limit,whateverthatlimitmay be. To aid in clinical interpreta-tion, thevariationin age-90(50% range)isreported as theinterval expectedto en-compass 50% of age-90 in the popula-tion. Positivecorrelations between limitand age-90 suggest that there is a ten-dencyfor children with lower motor de-velopment potential to reach their limitmore quickly (ie, have a lower age-90)

    Figure 2. Observed and Predicted Gross Motor Function Measure-66 (GMFM-66) Scores in Each Level of the Gross Motor FunctionClassification System

    100

    30

    60

    50

    40

    70

    80

    90

    20

    10

    0

    GMFM-66Score

    Level I Level II

    100

    30

    60

    50

    40

    70

    80

    90

    20

    10

    0

    0 1 152 3 4 5 6 7 8 9 10 11 12 13 14

    Age, y

    GMFM-66Score

    Level IV

    0 1 152 3 4 5 6 7 8 9 10 11 12 13 14

    Age, y

    0 1 152 3 4 5 6 7 8 9 10 11 12 13 14

    Age, y

    0 1 152 3 4 5 6 7 8 9 10 11 12 13 14

    Age, y

    Level V

    0 1 152 3 4 5 6 7 8 9 10 11 12 13 14

    Age, y

    Level III

    The curved solid lines indicate average performance. The horizontal dotted lines on the right of the figures indicate the band expected to encompass 50% of children slimits of development. The solid vertical lines indicate the average age-90. The dotted vertical lines indicate the bands expected to encompass 50% of age-90 valuesaround the average. The absence of 50% bands in level IV and level V indicates low variation in age-90 values.

    PROGNOSIS FOR GROSS MOTOR FUNCTION IN CEREBRAL PALSY

    1360 JAMA, September 18, 2002Vol 288, No. 11 (Reprinted) 2002 American Medical Association. All rights reserved.

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    6/8

    than children with higher potentials,even within GMFCS levels.

    The estimates of the average patternsof motor development in each stratumand the degree of individual differencesaround them (Table 2) havestraightfor-

    wardclinical interpretationswhen com-bined with knowledge of childrens ini-tial GMFCS level.Thus, forexample, themodel predicts thatthe expected limit ofa childs potentialinlevel III is54.3pointson theGMFM-66 (ie, thelevel IIImean),with 50% of childrens limits being be-tween 48.5 and 60.0 points. In terms oftherate of development,children inlevelIII are expected to have reached about90% of their potential by about age 3.7years. The positive correlation betweenlimit and age-90 for children in level IIIsuggests that a young child, who is per-

    forming at a higher level than expectedonthe basis ofthe average level III curve,is likely to level off sooner than his/herpeers. A substantial amount of prognos-tic information can thus be derived onthe basis of a single GMFM-66 assess-ment. The model incorporates possibleclassification errors within the GMFCSbecause these findings are based on chil-drens initial classification with no ef-fort to verifythe childs true level if thathappens to have changed over time.

    The residual SDs in Table 2 provide

    an indication of the degree to which themodelfitsforeachGMFCSlevel,andarea measure of how much each childsGMFM-66 scorecanbe expectedto varyaroundtheirtrueabilityover time.There

    is a suggestion in Table 2 that SDs fromthe model predictions are larger in lev-els I and V than the middle levels. TherawresidualSDswereplottedagainst pre-dicted values andagainstchildrensagesin each GMFCS stratum to address the

    adequacy of the modelfit. This revealedno tendency for the model errors to besystematically related to predicted valueor age. The residual SDs in Table 2 sug-gest that thesize of theexpectedwithin-child errors may be related to theGMFCS,whichsupports theuseof sepa-rate models for each stratum.

    To illustrate the clinical interpreta-tion ofthese curves,4 selected GMFM-66items have been identified on the ordi-nate ofthecurves (Figure3). TheGMFMitem 21 (diamond A) assesses whethera child canlift andmaintain his/her head

    in a vertical position with trunk sup-port by a therapist while sitting. A childwith a GMFM-66 score of 16 would beexpectedto have a 50% chanceof achiev-ing this task.8 This is something thatwould be seen relatively early in lifeamong children in GMFCS levels Ithrough IV, and only (on average) atabout age 2 years in children in level V.TheGMFM-66 item 24 (diamond B) as-sesses whether when in a sitting posi-tion on a mat, a child can maintain sit-ting unsupported by his/her arms for 3

    seconds. Childrenwould be expected tohave a 50% chance of being successfulatthis task at an averageGMFM-66 scoreof 32 points. This task would be rela-tively easily achieved by children in

    GMFCS levels I through III, much laterinchildren inlevel IV, and rarely by chil-dren in level V. The GMFM-66 item 69(diamond C) measures a childs abilityto walk forward 10 steps unsupported,a task associated with a mean GMFM-66

    scoreof56, andachievable (50%chance)predominantly by children in GMFCSlevels I and II. Finally, the task of walk-ing down 4 steps alternating feet with

    Figure 3. Predicted Average Developmentby the Gross Motor FunctionClassification System Levels

    100

    30

    60

    50

    40

    70

    80

    90

    20

    10

    0

    0 1 3 4 5 6 7 8 9 10 11 12 13 14 152

    Age, y

    GMFM-66

    Score

    D

    Level I

    Level II

    Level III

    Level IV

    Level V

    C

    B

    A

    Thediamonds on thevertical axisidentify 4 Gross Mo-tor FunctionMeasure-66 (GMFM-66)items that pre-dictwhenchildren areexpected to have a 50%chanceof completing that item successfully. The GMFM-66item 21 (diamond A) assesses whether a child can liftand maintain his/her head in a vertical position with

    trunk supportby a therapist while sitting;item 24 (dia-mond B) assesses whether when in a sitting positionon a mat, a child canmaintainsitting unsupported byhis/her arms for3 seconds;item69 (diamondC) mea-sures a childs abilityto walk forward 10 steps unsup-ported; anditem 87 (diamond D) assesses thetask ofwalking down 4 steps alternating feet witharms free.

    Table 2. Parameters of Motor Development for Gross Motor Function Classification System (GMFCS) *

    GMFCS Level

    I(n = 183)

    II(n = 80)

    III(n = 122)

    IV(n = 137)

    V(n = 135)

    Mean No. of observations per child 4.0 4.4 4.1 3.9 3.8

    GMFM-66 limit 87.7 68.4 54.3 40.4 22.3

    95% CI 86.0-89.3 65.5-71.2 52.6-55.8 39.1-41.7 20.7-24.0

    50% range 80.1-92.8 59.6-76.1 48.5-60.0 35.6-45.4 16.6-29.2

    Age-90, y 4.8 4.4 3.7 3.5 2.7

    95% CI 4.4-5.2 3.8-5.0 3.2-4.3 3.2-4.0 2.0-3.7

    50% range 4.0-5.8 3.3-5.8 2.5-5.5 3.5 2.7

    GMFM-66 limit/age-90 correlations 0.38 0.75 0.73 NA NA

    Residual SDs 3.9 2.8 2.0 2.4 3.1

    *GMFM-66 indicates Gross Motor Function Measure-66; CI, confidence interval; and NA, data not available because the variation in age-90 is near zero.Age-90 is the age in years at which children are expected to achieve 90% of their potential GMFM-66 score.The variation in age-90 was near zero, so the 50% range is approximately equal to the population average.

    PROGNOSIS FOR GROSS MOTOR FUNCTION IN CEREBRAL PALSY

    2002 American Medical Association. All rights reserved. (Reprinted) JAMA, September 18, 2002Vol 288, No. 11 1361

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    7/8

    arms free, which is GMFM-66 item 87(diamond D), will be observed at an av-erage GMFM-66 score of 81 points, andprobably only by children in GMFCSlevel I.

    COMMENT

    The patterns of motor development inchildren with cerebral palsy are the firstto be based on longitudinal observa-tions. They were created using a validclassification system of functional abili-ties and limitations of children with ce-rebral palsy and a systematic evalua-tion of gross motor function with anevaluative clinical instrument (GMFM).Data were collected from a large strati-fied random sample of children diag-nosedashavingcerebralpalsy, whowerereceiving a range of accepted medical,

    orthopedic, and developmental therapyservices. We believe the sample is rep-resentative of the population of chil-dren with cerebral palsy in Ontario, withresults generalizable to populationselse-where receiving similar types of mixeddevelopmental therapies.

    It is not clear whether children whoare currently receiving newer therapeu-tic modalities (selective dorsal rhi-zotomy,10 botulinum toxin,11,12 intra-thecal baclofen13 ) might performsubstantially better thanthe childrenin-

    volved in this study, and if so, whetherthis would limit the generalizability ofour findings. It should be noted that allof these recent therapeutic innovationsare used only with highly selected sub-groups of children with cerebral palsy,andthat even thebest results apply onlyto those specific groups and not to thewhole population.

    For example, following selective dor-sal rhizotomy, the reported improve-ments in gross motor function are sta-tistically significantly greater than those

    seen with physical therapy alone,buttheactual measured GMFM-66 changes arestill quite modest(mean measuredaddedbenefitinGMFM-66 changescoresin the[selectivedorsal rhizotomy plustherapy]group was 2.6)23 and are unlikely to beassociatedwith a changeofGMFCSlevel.Similarly, while the effects of botuli-numtoxin injections have been well de-

    scribed,11,12 the mean measured changeon the GMFM-88 was in the range ofabout 3%.Intrathecal baclofenis increas-ingly being used for the management ofspasticity in individuals with cerebralpalsy.Although individualshave shown

    improvements in spasticity andpain re-lief after receiving intrathecal baclofen,its effect on measured change in grossmotor function is limited.24 These com-ments are in no way meant to minimizethe effectiveness of these interventionsfor specific subgroups of children withcerebral palsy, but rather to note that atthis time these interventions do not, onaverage, have a major impact on func-tion as assessed with the GMFM.

    We expect that the findings from ourstudy will help parents understand theoutlook for their childs gross motor

    function, because an evidence-basedes-timatecannowbe made about gross mo-tor prognosisbased on age and GMFCSlevel.Thedata shouldprove equally use-ful to clinicians planning interventions,enabling clinicians andparents to makeinformed decisions about the most ap-propriatetherapy goalsforchildren.Thecurves also provide an effective way toassess whether a childs motor progressis consistent with patterns observed inchildren of similar age and severity.

    Because the GMFM-66assessments of

    children reported here were specificallymade without the use of aids, such aswalkers or crutches, these patterns ofgross motor developmentprobably rep-resent the lower limit of what childrenin each level can, on average, accom-plish in gross motor function. Further-more, the curves appear to reach pla-teaus by about age7 years. Children, onaverage, reach about 90% of their mo-tor function(as measured bytheGMFM-66) by around age 5 years or younger,depending on their GMFCS level.

    However, the curves reveal nothingabout the quality of motor control usedto accomplish the activities, which is anaspect of motor development that ap-pears to emerge later in childhood.25 Nordo the curves show how children applytheir motorfunction in thecontext ofac-tivity or participation in daily life,as for-mulated in the World Health Organiza-

    tions recent International Classificationof Functioning, Disability and Healthmodel.26 Furthermore,theGMFM-66as-sesses observed independent achieve-ment of motor function tasks, but doesnot (at least in this study) attempt to

    evaluate the ways in which childrensfunction is performed with or might beenhanced through the addition of aug-mentative and technical interventionssuch as aids, orthoses, or the use of pow-ered mobility to increase day-to-day in-dependence.27 Children maychange andimprove their gross motor performanceover the developing years through in-creased balance, stamina, energy effi-ciency, and quality of motor controlall featuresthat areimportantandshouldbe evaluated, but are beyond the scopeof the GMFM-66.

    Thus, it is extremely important thatparents, physicians, therapists, pro-gram managers,third-party payers, andother decision makers not assume fur-ther therapy is unhelpful or unneces-sary when thecurvesappear to level off.Continuing efforts shouldbe madeto ad-dress ways both to increase indepen-dent activity and to promote participa-tion of children with disabilities, as wellas to addresssecondary impairmentsthatmay arise. It should also be remem-bered that the children in the present

    study were receivinga range of contem-porary developmental therapy servicesthat we believe are representative of thetherapiesprovidedin theWesternworld.It is likely that as new therapiesemerge,patterns of motor development in chil-dren diagnosedas having cerebral palsymay change and modifications to thesemodels will be needed. We believe thatthe motor development curves will haveimportant applications for the evalua-tion of specific interventions by permit-ting analysisof theextent to which these

    interventions improvea childs grossmo-tor function beyond what is predictedbased on age and GMFCS level.

    Based on our previous work, an as-sumption of this study was that GMFCSlevelsare stable over time, making prog-nostication meaningful. Wood andRosenbaum15 demonstrated an overallreliability ofGMFCS over time (fromage

    PROGNOSIS FOR GROSS MOTOR FUNCTION IN CEREBRAL PALSY

    1362 JAMA, September 18, 2002Vol 288, No. 11 (Reprinted) 2002 American Medical Association. All rights reserved.

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/
  • 8/3/2019 PrognosisforGrossMotorFunctionMGCs

    8/8

    2 years to 12 years) of 0.79, withhigher valueswhen onetrackedthecon-sistency of GMFCS levels from age 2 to4 years to age 12 years (0.82) or fromage 4 to 6 years to age 12 years (0.87).Children in our study were allocated to

    a GMFCS level at entry tothe study,anddatawere analyzed according to that ini-tial assignment. This was done to re-flect theclinicalreality that parents seekprognostication about their childs out-look from the time the disability is firstdiagnosed. As development progresses,children may be reclassified. If this oc-curs, it is important for clinicians to re-formulate the prognosis, based on themost recent assessment of a childs mo-tor activities and GMFCS stratum.

    The present data can be used to ex-plore the creation of motor growthcurves for children with different dis-tributions of cerebral palsy, as has beendone by others.28 These analyses arecurrently under way. However, there isreason to believe that in the absence ofa systematic, protocol-driven classifi-cation of the topographical compo-nents of cerebral palsy, reliability ofsuch categorizations is relatively poor.29

    For the same reason, we have not yetanalyzed the findings according to theformof motor impairment (spastic, dys-tonic, ataxic, or mixed),which hasbeen

    described elsewhere.30 In fact, approxi-mately76% of the children in this studywere described as having spastic cere-bral palsy, with much smaller num-bers in the other subgroups.

    The curves describe patterns forgroups of children. There is within-stratum variation in motor develop-ment, which is based on other aspectsofeach childs functional status.31,32 Moreresearch is needed to understand, andto be able to measure accurately, theim-pact of factors such as a childs visual

    ability, cognitive capacity, motivation,parental encouragement, and the con-tribution of therapies that might be as-sociated with individual variation inprogress.It will also beimportant to con-tinue to follow the motor developmentof these children through adolescence,because much remains to be learned

    about the impact of puberty and the de-mands of secondary school on motorfunction andactivity of adolescents withcerebral palsy.Finally, we expectthat re-searchby our group andothers will pro-vide validation of the accuracy and util-

    ity of these curves.Author Contributions: Study concept and design:Rosenbaum, Walter, Palisano, Russell, Wood.Acquisition of data: Rosenbaum, Walter, Russell, Raina,Galuppi.Analysis and interpretation of data: Rosenbaum,Walter, Hanna, Palisano, Raina, Bartlett.Drafting of the manuscript: Rosenbaum, Walter,Hanna, Palisano, Raina.Critical revision of the manuscript for important in-tellectual content: Rosenbaum, Walter, Hanna,Palisano, Russell, Raina, Wood, Bartlett, Galuppi.Statistical expertise: Walter, Hanna, Raina.Obtained funding: Rosenbaum, Walter, Palisano,Russell.Administrative, technical, or material support:Rosenbaum, Palisano, Raina, Bartlett, Galuppi.Study supervision: Rosenbaum, Russell.Funding/Support: This study was made possible bygrant MT-13476from the CanadianInstitutes of HealthResearch (CIHR)(formerly the Medical Research Coun-cil of Canada) andgrant R01-HD-34947from the Na-tional Center for Medical Rehabilitation Research oftheNationalInstitute of Child Healthand Human De-velopment.PreviousPresentation: A preliminary reportof thefind-ings of thisstudywas presented at theEuropeanAcad-emy for Childhood Disability, Goteborg, Sweden,October 13, 2001.Acknowledgment: We thank the children and fami-lies, whose ongoingparticipationmade thisstudy pos-sible. We also acknowledge the considerable contri-butions of thecoordinatorsof thestudyat eachcenter,and thetireless work ofmore than 125 therapists whoassessed children during the course of this study.

    REFERENCES

    1. Stanley FJ, Blair E, Alberman E. Cerebral Palsies:Epidemiology & Causal Pathways. London, England:Mac Keith Press; 2000:29.2. Mutch LW, Alberman E, Hagberg B, Kodama K,Velickovic MV. Cerebralpalsy epidemiology.DevMedChild Neurol. 1992;34:547-555.3. Bleck EE. Locomotor prognosis in cerebral palsy.Dev Med Child Neurol. 1975;17:18-25.4. Molnar GE, Gordon SV. Predictive value of clinicalsignsfor earlyprognosticationof motorfunction in cere-bral palsy. Arch Phys Med Rehabil. 1974;57:153-158.5. Watt JM, Robertson CM, Grace MG. Early prog-nosis for ambulation of neonatal intensive care sur-vivors with cerebral palsy. Dev Med Child Neurol.1989;31:766-773.6. Crothers B, Paine RS. Natural History of CerebralPalsy. Cambridge, Mass: HarvardUniversity Press;1959.7. Scrutton D, RosenbaumPL. Thelocomotor devel-opment of children with cerebral palsy. In: Connolly

    K, ForssbergH, eds. Neurophysiology and Neuropsy-chology of Motor Development. London: Mac KeithPress; 1997:101-123.8. Palisano RJ, Hanna SE, Rosenbaum PL, et al. Vali-dation of a model of gross motor function for childrenwith cerebral palsy. Phys Ther. 2000;80:974-985.9. Palisano RJ,Rosenbaum PL,Walter SD,Russell DJ,Wood EP,Galuppi BE.Developmentand reliability ofa system to classify gross motor function in childrenwith cerebral palsy. DevMed Child Neurol. 1997;39:214-223.

    10. Steinbok P. Outcomes after selective dorsal rhi-zotomy for spastic cerebral palsy. Childs Nerv Syst.2001;17:1-18.11. Mall V, Heinen F, Kirschner J, et al. Evaluation ofbotulinum toxin. J Child Neurol. 2000;15:214-217.12. Kirschner J, Linder M, Berweck S, et al. Treat-ment of pes equines in cerebral palsy with botulinumtoxin A andfunctionalbenefit accordingto gross mo-tor function measure. Eur J Neurol. In press.

    13. Almeida GL, Campbell SK, Girolami G, Penn R,Corcos DM. Multidimensional assessment of motorfunction in a child withcerebral palsy followingintra-thecal administration of baclofen. Phys Ther. 1997;77:751-764.14. Collet JP,Vanasse M, Marois P, et al.Hyperbaricoxygen for children with cerebral palsy. Lancet. 2001;357:582-586.15. Wood EP,RosenbaumPL. TheGross Motor Func-tionClassification System for cerebral palsy. DevMedChild Neurol. 2000;42:292-296.16. Russell DJ,Rosenbaum PL, Cadman DT,et al.Thegross motor function measure. Dev Med Child Neu-rol. 1989;31:341-352.17. Russell D, Palisano R, Walter S, et al. Evaluatingmotor functionin children with Downsyndrome. DevMed Child Neurol. 1998;40:693-701.18. Russell DJ,Avery L,Rosenbaum PR,Raina P,Wal-terSD, Palisano RJ.Improvedscalingof thegross mo-tor function measure for children with cerebral palsy.Phys Ther. 2000;80:873-885.19. RussellDJ, Rosenbaum PL,AveryL, Lane M. GrossMotor Function Measure (GMFM-66 and GMFM-88) Users Manual: Clinics in Developmental Medi-cine,No. 159.London,England:Mac KeithPress; 2002.20. Avery L, Russell D, Raina P, Rosenbaum P, Wal-terS. Rasch analysis of thegross motorfunctionmea-sure. Arch Phys Med Rehabil. In press.21. Russell DJ, Rosenbaum PL, Lane M, et al. Train-ing users in the gross motor function measure. PhysTher. 1994;74:630-636.22. Pinheiro JC, Bates DM. Mixed Effects Models inS and S-PLUS. New York, NY: Springer; 2000.23. McLaughlin J, Bjornson K, Temkin N, et al. Se-lective dorsal rhizotomy: meta-analysis of three ran-domized controlledtrials. DevMed ChildNeurol.2002;44:17-25.24. Butler C, Campbell S. Evidence of the effects ofintrathecal baclofen for spastic and dystonic cerebral

    palsy: AACPDM Treatment Outcomes Committee Re-view Panel. Dev Med Child Neurol. 2000;42:634-645.25. Boyce W, Gowland C, Rosenbaum P, et al. Thegross motor performance measure: validity and re-sponsiveness of a measure of quality of movement.Phys Ther. 1995;75:603-613.26. World Health Organization. International Classi-fication of Functioning, Disability and Health.Geneva,Switzerland: World Health Organization; 2001. Avail-able at: http://www3.who.int/icf/icftemplate.cfm. Ac-cessibility verified August 26, 2002.27. Butler C. Augmentative mobility: why do it? PhysMed Rehabil Clin North Am. 1991;2:801-816.28. Beckung E, Carlsson G, Carlsdottor, Uvebrant P.The natural history of gross motor function in chil-drenwith cerebralpalsy in western Sweden. DevMedChild Neurol. 2001;43(suppl 89):27.29. Blair E, Stanley F. Interobserver agreementin the

    classsification of cerebral palsy. Dev Med Child Neu-rol. 1985;27:615-622.30. Cans C. Surveillance of cerebral palsy in Europe:a collaboration of cerebral palsy surveys and regis-ters. Dev Med Child Neurol. 2000;42:816-824.31. KennesJ, RosenbaumPL, HannaSE, et al.Healthstatus of school-agedchildren withcerebral palsy. DevMed Child Neurol. 2002;44:240-247.32. Bartlett DJ, Palisano RJ. A multivariate model ofdeterminants of motor change for children with ce-rebral palsy. Phys Ther. 2000;80:598-614.

    PROGNOSIS FOR GROSS MOTOR FUNCTION IN CEREBRAL PALSY

    2002 American Medical Association. All rights reserved. (Reprinted) JAMA, September 18, 2002Vol 288, No. 11 1363

    at McMaster University on February 22, 2009www.jama.comDownloaded from

    http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/http://jama.ama-assn.org/