Limb distribution, The term cerebral palsy or CP (more appropriately the cere-

bral palsies) refers to a group of disorders in the development

of postural control and mobility secondary to non-progres-
motor impairment, and sive impairments of the developing central nervous system
(Bax 1964, Mutch et al. 1992, Stanley et al. 2000). It is recog-
functional classification nized that many developmental disorders, notably a number
of syndromes including mental retardation* or develop-

of cerebral palsy mental delay, include motor impairment but can at times be
described and classified more usefully as disease entities in
other ways (Badawi et al. 1998). Nonetheless, the idea of CP
as a group of developmental disorders of motor control is
Jan Willem Gorter MD PhD, Physician in Physical Medicine thought to be important and useful as both a clinical and an
and Rehabilitation, Rehabilitation Centre De Hoogstraat, epidemiological concept.
Utrecht, the Netherlands. The history of approaches to classification of CP has been
Peter L Rosenbaum* MD FRCP(C), Professor of Paediatrics; presented by Ingram (1984). Both there and in subsequent
Steven E Hanna PhD, Assistant Professor, Department of work (Stanley et al. 2000) the traditional systems of descrip-
Clinical Epidemiology and Biostatistics, McMaster tive classification based on impairments have been well out-
University, Canada. lined. These systems include an account of the topography
Robert J Palisano ScD, Professor, Programs in Rehabilitation of CP (what parts of the body are affected), the types of motor
Sciences, Drexel University, Philadelphia, PA, USA. impairment (describing the predominant characteristics of the
Doreen J Bartlett PhD, Assistant Professor, School of Physical motor findings), and the severity of motor impairments (Balf
Therapy, Faculty of Health Sciences, University of Western and Ingram 1955). Others have tried to classify cerebral palsies
Ontario, London, Ontario; on the basis of pathological findings (as outlined by Ingram
Dianne J Russell MSc, Associate Professor, School of 1984) and more recently by cerebral imaging techniques (Pinto-
Rehabilitation Science; Martin et al. 1995). The recent modification of the World Health
Stephen D Walter PhD, Professor; Organizations (2001) conceptual framework about health con-
Parminder Raina PhD, Associate Professor, Department of ditions and functioning, the International Classification of
Clinical Epidemiology and Biostatistics, McMaster University; Functioning, Disability and Health, provides another useful
Barbara E Galuppi BA, Project Coordinator, Ontario Motor way of considering CP and its consequences, from the perspec-
Growth Study, CanChild Centre for Childhood Disability tives of biological factors (impairments), functional impacts
Research, Hamilton, Ontario; (activity limitations), and the social consequences of the con-
Ellen Wood MD FRCP(C) MSc, Assistant Professor, Faculty of dition (participation restrictions).
Medicine, Dalhousie University, Halifax, Nova Scotia, Canada. To understand the clinical picture of CP we need to know
the value of characteristics at the impairment level, such as
*Correspondence to second author at CanChild Centre for the limb distribution of the clinical syndromes (the number
Childhood Disability Research, IAHS, Room 408, 1400 Main of limbs with impaired motor control) or the type of motor
Street West, Hamilton, ON, Canada, L8S 1C7. disorder, and its severity at the function level. The primary
E-mail: [email protected] purpose of this report was to describe how limb distribution
and type of motor impairment (spastic, dyskinetic, ataxic, or
other) relate to functional abilities described by the Gross
Motor Function Classification System (GMFCS). The second
purpose was to explore to what extent patterns of motor
This study explored the relationships between the Gross Motor development of children with CP can be explained by the
Function Classification System (GMFCS), limb distribution, and limb distribution of CP and by type of motor impairment, in
type of motor impairment. Data used were collected in the contrast to observations made using the GMFCS alone.
Ontario Motor Growth study, a longitudinal cohort study with a
population-based sample of children with cerebral palsy (CP) in BACKGROUND ISSUES
Canada (n=657; age 1 to 13 years at study onset). The majority One of the continuing challenges in the field of the cerebral
(87.8%) of children with hemiplegia were classified as level I. palsies concerns what aspects of these conditions to classify,
Children with a bilateral syndrome were represented in all GMFCS and how to do so. Classification can serve one or more of sev-
levels, with most in levels III, IV, and V. Classifications by eral purposes (Alberman 1984), and the system(s) used should
GMFCS and limb distribution or by GMFCS and type of motor be specific to those aims. Epidemiologists want to track the
impairment were statistically significantly associated (Pearsons incidence, prevalence, and features of these conditions over
2 p<0.001), though the correlation for limb distribution (two time to ascertain whether and how these indices are chang-
categories) by GMFCS was low (tau-b=0.43). An analysis of ing (see Krgeloh-Mann et al. 1993, Blair and Stanley 1997,
function (GMFCS) by impairment (limb distribution) indicates Hagberg et al. 2001). This requires clinical descriptions at the
that the latter clinical characteristic does not add prognostic impairment level of both primary features, such as limb distri-
value over GMFCS. Although classification of CP by impairment bution and type of motor impairment, as well as associated
level is useful for clinical and epidemiological purposes, the value features of the conditions (such as epilepsy). Parents and fam-
of these subgroups as an indicator of mobility is limited in ilies wish to have an account of the severity of the condition
comparison with the classification of severity with the GMFCS. and to understand the prognosis of their childs mobility, for
See last page for list of abbreviations. *UK usage: learning disability.

Developmental Medicine & Child Neurology 2004, 46: 461467 461

which functional and prognostic classification are necessary (Jarvis and Hey 1984). Several reports from CP registries
(Palisano et al. 2000, Wood and Rosenbaum 2000, Rosenbaum (Krgeloh-Mann et al. 1993, Pharoah et al. 1998, Stanley et al.
et al. 2002). Clinical service providers and planners have to be 2000, Beckung and Hagberg 2002) have illustrated the correla-
able to describe their populations in ways that can be used for tions between motor severity and other aspects of neurode-
service planning (e.g. for creation of a spasticity management velopmental impairment, such as epilepsy, mental retardation,
clinic) for which descriptions of both functional status and and sensory impairments. These findings suggest that sever-
comorbidity are important. Such information also informs ity is a useful marker of at least some aspects of the clini-
individual childrens and families needs for issues such as cal picture. However Kennes et al. (2002) found that, apart
caregiver assistance, equipment, and services. from ambulation (tau-b=0.82) and dexterity (tau-b=0.58),
Very little evidence is available about the reliability of exist- the correlations between a valid marker of motor severity
ing classifications of children with CP, and what has been pub- (GMFCS level) and other dimensions of functional status (e.g.
lished suggests that classification of the clinical features of CP speech or sensory function) were at best modest (tau-b=0.46
is difficult and of relatively poor reliability (Table I). In one of and 0.36 respectively) and often statistically non-significant.
the rare studies to explore this challenge, Blair and Stanley The prognostic validity of the GMFCS has been reported
(1985) reported that even among a group of six experienced with the use of both cross-sectional (Palisano et al. 2000) and
neurodevelopmental clinicians assessing children with known longitudinal (Rosenbaum et al. 2002) observations of the
CP, it was difficult to reach as high as 60% agreement on sever- gross motor development of children with CP followed over
ity of the disability, and lower rates of agreement were observed several years in the Ontario Motor Growth study. Serial
regarding type of motor impairment (40%) and body location assessments of motor function of 657 children with CP were
(50%). To our knowledge (E Blair, personal communication; made with a reliable and valid measure: the Gross Motor
MA Johnson, personal communication) no other work than Function Measure (GMFM; Russell et al. 1989). At the same
that described in Table I has been published to assess inter- time a clinical description of each childs CP was obtained,
observer agreement about the classification of CP according including both the limb distribution of the CP and the pre-
to type of motor impairment and the degree of limb involve- dominant type of motor impairment. This provided the oppor-
ment. Palisano et al. (1997) and Wood and Rosenbaum (2000) tunity to explore relationships between the various clinical
have both reported good to excellent interrater reliability for descriptions of children with CP.
severity of gross motor function limitations in children with
CP using the GMFCS. One might add that if a system cannot Methods
be used reliably it is not possible for it to be valid (i.e. to Details of the Ontario Motor Growth study have been reported
reflect what it is meant to reflect) because of uncertainty elsewhere (Rosenbaum et al. 2002) and are presented only
about the accuracy of the categorizations. briefly here. This study was made possible through a partner-
The validity of a classification system rests essentially on ship between the CanChild Centre for Childhood Disability
the usefulness of its categories. In other words, do they enable Research at McMaster University and the 19 publicly-funded
people to make meaningful distinctions between the sub- regional ambulatory childrens rehabilitation programs in
groups? In the field of CP one of the challenges has been the Ontario, Canada. Each program serves the majority of eligible
need to strike a balance between the varied nature of the clin- children in their area.
ical entities that constitute the CP spectrum and the com- The ethics review boards of Hamilton Health Sciences
plexity of systems to classify these multivariable conditions Corporation, the Bloorview MacMillan Childrens Centre

Table I: Summary of literature on reliability and validity of classification systems in cerebral palsy

Classification References ICF level Reliability Validity

Type of motor impairmenta Blair and Stanley (1985) Impairment

Surveillance of Cerebral Palsy in Europe (2000) Impairment

Limb distributionb Blair and Stanley (1985) Impairment

Krgeloh-Mann et al. (1993) Impairment
Surveillance of Cerebral Palsy in Europe (2000) Impairment
Severity
Mild/moderate/severe Balf and Ingram (1955) Impairment/activity
Mild/moderate/severe Blair and Stanley (1985) Not explicitly described
Four functional gradingsc Krgeloh-Mann et al. (1993) Impairment/activity
GMFCS (five levels) Palisano et al. (1997); Wood and Rosenbaum (2000) Activity + +
ICIDH Handicap coded Beckung and Hagberg (2000) Activity/participation
aFor example, spastic, dyskinetic, ataxic, hypotonic, or mixed. bFor example, hemiplegia, diplegia, triplegia, or quadriplegia.
cDerived from the classification system of Hagberg et al. (1975).dDimension mobility, 10 levels.

Scoring: +, classification has been studied systematically and meets criteria of good reliability and/or evidence of validity; , reliability/validity
has been studied systematically, but has not been fully established; , classification has not been tested or information is unavailable.
ICF, International Classification of Functioning, Disability and Health; GMFCS, Gross Motor Function Classification System;
ICIDH, International Classification of Impairment, Disability and Handicap.

462 Developmental Medicine & Child Neurology 2004, 46: 461467

(Toronto, Ontario), and the Thames Valley Childrens Centre of health status (Kennes et al. 2002) no standardizd informa-
(London, Ontario) approved the Ontario Motor Growth tion was obtained on cognitive functioning (e.g. IQ), visual
study (Rosenbaum et al. 2002). All parents of the participants impairments (field defects), or comorbidity (e.g. epilepsy).
gave their written consent.
The sampling frame was created in early 1996 with 18 of MEASURES
the 19 centres and one hospital-based therapy program in a Severity of CP was based solely on gross motor function as
community without a regional centre. Each centre identified judged by therapists using the GMFCS, a reliable and valid
all the children on their case list with a diagnosis of CP born five-level pattern recognition classification system that dis-
in 1986 or later. In addition, children with neuromotor find- criminates between children with CP according to their age-
ings consistent with CP (e.g. spasticity or reflex abnormalities) specific gross motor activity (Palisano et al. 1997, Wood and
who had not yet been diagnosed as having cerebral palsy were Rosenbaum 2000). The GMFCS describes the major func-
included. The sampling frame contained 2108 children, of tional characteristics of children with CP in each level within
whom 1304 were randomly selected. The goal was to obtain several age windows: before their second birthday, between
a random sample of eligible children, stratified into four age age 2 years and the fourth birthday, between age 4 years and
groups and five GMFCS severity levels. Each age/ GMFCS stra- the sixth birthday, and between ages 6 and 12 years. Table II
tum was deliberately over-sampled in the hope of getting at outlines the main abilities of children aged 6 to 12 years in
least 15 children in each predefined stratum. In all, 365 chil- each GMFCS level. Use of the GMFCS requires familiarity
dren were ineligible or unavailable for various reasons. Of the with the child but requires no formal training.
remaining 939 children, 721 families (77%) consented and, of Motor function was assessed with the 88-item version of
these, 682 (94.6%) provided data. In total, 657 had fully use- the GMFM (Russell et al. 1989) and subsequently analyzed
able data (369 males [56.2%], 288 females [43.8%]), after the using the Gross Motor Ability Estimator computer scoring
exclusion of children who were subsequently determined program to get an interval-level GMFM-66 score (Russell et
not to have CP. The final sample included 183 children in al. 2000, 2002). The GMFM is a widely used criterion-refer-
GMFCS level I, 80 in level II, 122 in level III, 137 in level IV, and enced clinical observation tool developed for and validated
135 in level V. Mean age of the children at the start of the study on children with CP. It was not designed to compare the func-
was 6.6 years (SD 2.8); no significant difference was found for tion of children with CP to typically developing children. The
age within each GMFCS stratum. GMFM measures gross motor function in lying and rolling,
At the first assessment, therapists were asked in a standard- crawling and kneeling, sitting, standing, and walk-run-jump
ized way whether the child had received a formal diagnosis of activities and can be used with any child or adolescent with
CP. Therapists were asked to report the limb distribution of the CP. The original 88-item measure has excellent reliability and
childs CP as it was described in the childs clinic chart by indi- a demonstrated ability to evaluate meaningful change in
cating the predominantly affected limbs in a figure with the gross motor function in children with CP (Russell et al. 1989),
clinical subtypes of bilateral CP developed by Michaelis and as does the newer 66-item GMFM (Russell et al. 2000).
Edebol-Tysk (1989) and also published by Krgeloh-Mann et
al. (1993). No specific instructions were given to the clinician ANALYSIS
to define hemiplegia, diplegia, triplegia, or quadriplegia, nor Descriptive analyses were done to report the association
was a formal algorithm available at the time by which to stan- between the three methods of classification of CP, i.e. (a) by
dardize the therapists reports. limb distribution (two categories [one-sided vs two-sided
Therapists were also asked to include whatever terms had involvement], binary/ordinal system, and four categories
been used to describe the diagnosis, including terms about the [hemiplegia, diplegia, triplegia, quadriplegia], nominal sys-
type of motor impairment (spastic, dyskinetic, ataxic, hypoton- tem); (b) motor type (four categories, nominal system); and
ic). When no formal diagnosis had yet been given, therapists (c) severity by GMFCS (ordinal ranking). To test the statistical
were asked whether, in their judgement, that childs motor significance and, where applicable, to quantify the degree of
behaviour and patterns looked like CP. Information on the association between the classification systems, Pearsons 2
limb distribution of CP was missing for 17 (2.8%) children, test and Kendall tau-b were used respectively. A correlation
and information on the type of motor impairment was miss- (tau-b) below 0.7 is considered to be poor to modest, because
ing for 18. For the purposes of the present study, children such an association accounts for, at most, about 49% of the
with three- and four-limb CP were grouped together (n=325). explained variance (i.e. tau-b2).
In the Ontario Motor Growth study therapists were asked to Non-linear mixed-effects modelling was used to estimate the
report any medical problem in the period 6 months before parameters of motor development as described in Rosenbaum
entry to the study, but apart from a parent-completed report et al. (2002) for children in each stratum by limb distribution

Table II: Summary account of gross motor function by GMFCS level at ages 6 to 12 years (Palisano et al. 1997)

Level Description

I Walks without restrictions; limitations in more advanced gross motor skills

II Walks without assistive devices; limitations walking outdoors and in the community
III Walks with assistive mobility devices; limitations walking outdoors and in the community
IV Self-mobility with limitations; children are transported or use power mobility outdoors and in the community
V Self-mobility is severely limited even with the use of assistive technology

Classification of Cerebral Palsy Jan Willem Gorter et al. 463

(hemiplegia, diplegia, and quadriplegia) and for children in Within each GMFCS level we compared the differences in
each of the five GMFCS levels. Motor development of children GMFM-66 limit scores between subgroups classified by limb
with CP on a group level was expressed as a non-linear change distribution, using their 95% CI. Because there were chil-
function with parameters that represent the mean rate of change dren described as having diplegic CP in each of GMFCS levels
and mean limit of observed GMFM-66 scores. We report the IIV, the diplegic group was used as the reference population
mean GMFM-66 limit scores as a value between 0 and 100. To against which to compare the patterns of motor development
aid interpretation, the mean rate of gross motor development in children with other distributions (using data reported by
is expressed as age-90, which is the mean age at which children Rosenbaum et al. 2002). In level V no comparison was possi-
with CP reach 90% of their predicted GMFM-66 limit. Ninety- ble, because only one child in that level was classified as having
five percent confidence intervals (CIs) are provided for both diplegic CP to contrast with 126 children having a quadri-
quantities. Estimated variances in the limit for each subgroup plegic distribution.
are used to construct intervals that are expected to encompass In the analysis, children with any type of motor impairment
50% of limits in the population; this is expressed as the 50% both spastic and other types of motor impairment were
range of the GMFM-66 limit. Likewise, the variation in age-90 included. We refrained from undertaking additional subgroup
(50% range) is reported as the interval expected to encompass analysis (e.g. looking at children with spastic CP only, visual
50% of the age-90 values around the mean age-90. impairments, cognitive capacity or comorbidity) within each

100 11.3 3.4

GMFCS level
5.7
Level I
36.9 11.3 13.7
80
% within limb distribution

Level II

24.2 Level III

31.2
60 87.8
23.5 Level IV

Level V
40
45.2
31.3 46.0
20
7.1
2.0 7.8 8.1
3.1 0.5
0
Hemiplegia (n=98) Diplegia (n=217) Triplegia (n=62) Quadriplegia (n=263)

Figure 1: Limb distribution by GMFCS; data from Ontario Motor Growth study (Rosenbaum et al.
2002). Kendalls tau-b (limb involvement in two categories: one-side versus two-side involvement by
GMFCS) 0.43, p<0.001; Kendalls tau-b (limb involvement in four categories: hemiplegia, diplegia,
triplegia, quadriplegia by GMFCS) 0.13, p=0.001. Pearsons 2 test (limb involvement in four categories:
hemiplegia, diplegia, triplegia, quadriplegia by GMFCS) p<0.001.

100 5.1 GMFCS level

12.5 6.9

12.8
Level I
30.8 15.5
% within motor impairment

33.0
80 Level II
15.4
12.0
50.0 11.5
Level III
60 11.0
Level IV
30.8
19.2 32.8
19.6
Level V
40
15.4
18.8
31.2
20 35.9 32.8
23.1
17.6
6.3
0
Spastic (n=500) Dyskinetic (n=39) Ataxic (n=16) Hypotonic (n=26) Mixed (n=58)

Figure 2: Distribution of type of motor impairment by GMFCS; data from Ontario Motor Growth study
(Rosenbaum et al. 2002). Pearsons 2 test (motor impairment by GMFCS) p<0.001.

464 Developmental Medicine & Child Neurology 2004, 46: 461467

GMFCS stratum, because this is possible only for variables There were significant differences in the mean limits of motor
recorded at the outset of the study in a systematic and reliable development by GMFCS strata, with no overlap in the 95%
way, issues that were not a focus of the original Ontario CIs between adjacent levels. Furthermore, the range of values
Motor Growth study. within the 95% CIs was relatively narrow (only in level II did
the value reach 5.7 GMFM-66 points). Corresponding values for
Results limits of gross motor development of children classified on
Figures 1 and 2 present the functional classification of the sam- the impairment level, namely hemiplegia, diplegia, and quadri-
ple (by GMFCS) with respect to the four limb distribution plegia, were significantly different from one another, because
groups (hemiplegia, diplegia, three- and four-limb CP) and by the 95% CIs did not overlap. Here, however, the width of the
the predominant type of motor impairment respectively. 95% CI range was larger: the minimum value was 5.7 GMFM-66
There was no significant difference in the mean ages of the points (diplegia) and the maximum was 15.1 GMFM-66 points
children in the four limb distribution groups or by type of for children classified as having quadriplegia. In addition, the
motor impairment. According to limb distribution most chil- 50% ranges of GMFM-66 scores were in most cases larger for
dren with hemiplegia were classified in GMFCS level I (87.8%) the limb distribution groups (three categories) than for the five
but a small number were classified in level II and very few in GMFCS groups, indicating that the estimated rate and limit
levels III or IV. Children with a bilateral syndrome were rep- varied substantially within subgroups of limb distribution.
resented in all GMFCS levels, with the majority in levels III, The rate of gross motor development has similar straight-
IV and V. The association between limb distribution and forward clinical interpretations when combined with the chil-
GMFCS levels was modest at best, and varied depending drens GMFCS level. Although statistically non-significant, the
on whether one classified using four categories of limb dis- age by which a child is expected to have reached about 90%
tribution (tau-b=0.13; p=0.001) or two categories of limb of their average limit has a positive correlation with the limit
distribution (tau-b=0.43; p<0.001). Both relationships, for GMFM-66 score for each GMFCS level, but not for each cate-
limb distribution (four categories) and type of motor impair- gory of limb distribution. Because almost 80% of the children
ment, were statistically significantly associated with func- were described as having spastic CP, no analysis of GMFM
tional motor ability described using the GMFCS (Pearsons 2 scores by type of motor impairment has been undertaken.
p<0.001). The mean limit scores of the reference group in each level
For children with presumed spastic CP only (n=501) the (diplegic CP in levels I to IV, quadriplegia in level V) with 95%
relationship between limb distribution and gross motor abil- CIs of the estimates are reported in Table IV. In addition, for
ity (GMFCS) was essentially the same as reported for chil- each subgroup a 50% range interval of the GMFM-66 limit
dren with all types of motor impairment (two categories of could be calculated. Analysis of function (GMFCS) by impair-
limb distribution, Kendall tau 0.46, p<0.001; four categories ment (limb distribution) was performed, to contrast patterns
of limb distribution, Kendall tau 0.10, p=0.02; Pearsons 2 of gross motor development in the predominant limb dis-
p<0.001). tribution groups for each level. It can be seen that the dif-
Table III presents the GMFM-66 limit and rate (age-90) ferences in mean GMFM limit scores between hemiplegia
estimates for the mean gross motor development according and diplegia in level I and diplegia versus quadriplegia in
to limb distribution and according to functional ability as level II were small (2.3 and 0.4 GMFM-66 points respective-
described by the GMFCS. As reported in Rosenbaum et al. ly) and not significant. In levels III and IV the children with
(2002), the estimated limit of development was highest in diplegic CP had small (3.6 or 3.7 GMFM-66 points) but sig-
GMFCS level I and lowest in level V: 87.7 and 22.3 respectively. nificantly higher GMFM limits than those with three or four

Table III: Parameters of motor development for severity (GMFCS levels I to V) and limb distribution

Category n Mean observations GMFM-66 95% CI 50% range Age-90 95% CI 50% range
per child Limit (years) (Age-90) (Age-90)

GMFCS*
I 183 4.0 87.7 86.089.3 80.192.8 4.8 4.45.2 4.05.8
II 80 4.4 68.4 65.571.2 59.676.1 4.4 3.85.0 3.35.8
III 122 4.1 54.3 52.655.8 48.560.0 3.7 3.24.3 32.55.5
IV 137 3.9 40.4 39.141.7 35.645.4 3.5 3.24.0 3.5b
V 135 3.8 22.3 20.724.0 16.629.2 2.7 2.03.7 2.7b
Limb distribution
Hemiplegia 98 4.2 87.9 82.991.7 78.893.5 4.6 3.65.9 3.56.1
Diplegia 217 4.0 72.3 69.375.0 57.072.3 4.6 4.25.1 3.56.1
Quadriplegiaa 325 3.9 38.3 31.046.1 24.054.9 3.4 2.74.3 2.64.5

*Data as reported in Rosenbaum et al. (Copyrighted (2002). American Medical Association. All rights reserved). Parameters of motor
development (limit, 95% confidence interval, 50% range) are expressed in GMFM-66 scores. Age-90 is the age at which children are expected to
achieve 90% of their potential GMFM-66 score.
aFor this analysis, data on all children with triplegia and quadriplegia were collapsed into quadriplegia group.
bVariation in age-90 was near zero, so 50% range is approximately equal to population mean.

CI, confidence interval; GMFCS, Gross Motor Function Classification System; GMFM, Gross Motor Function Measure.

Classification of Cerebral Palsy Jan Willem Gorter et al. 465

limbs involved (grouped as quadriplegia). (Palisano et al. 1997, Wood and Rosenbaum 2000).
The availability of a large database of information about
Discussion motor development in a community-based randomly select-
The clinical features of CP conventionally used to describe ed population of children with CP has made it possible to
both populations and individual children focus on the impair- explore the relative usefulness of impairment-level and activ-
ment level (i.e. limb distribution and type of motor impair- ity-level methods of classifying CP. In this study we found that
ment). Until recently, functional limitations were described the overall association between classification on the impairment
idiosyncratically, although with the development of the level (limb distribution and type of motor impairment) and
GMFCS a reliable and discriminative system can be applied to classification according to function (by GMFCS) is statistical-
this clinical dimension of CP. Of these approaches, only the ly significant but low. The GMFCS seems to provide meaning-
GMFCS has demonstrated validity and clinical utility. It is possi- ful distinctions in gross motor development between five
ble, in fact likely, that a multidimensional classification that functional subgroups (Rosenbaum et al. 2002). In contrast,
includes functional, topographical (limb distribution), and grouping by limb distribution or type of motor impairment
motor impairment information in combination with associat- does not provide the clinician additional prognostic infor-
ed conditions simultaneously for example, cognitive capac- mation in terms of gross motor abilities. Limb distribution
ity, visual impairments, and comorbidity (epilepsy) would may differentiate children only in levels III and IV, but not in
enable parents, service providers, researchers, epidemiolo- levels I, II, and V. Given the very small differences observed, it
gists, and others involved in CP to make more meaningful is not likely that this information will enable clinicians to dis-
distinctions about both prognostic and therapeutic issues tinguish children in ways that families might find useful for
within this heterogeneous population. prognostic or interventional planning.
Turning first to the issue of classification by topography In this study, with the exception of the GMFCS data, the
(limb distribution), we have been unable to find clear and clinical descriptions of children with CP were made by clini-
meaningful descriptions of the distinctions between, for exam- cians without the benefit of standardized information or a
ple, severe diplegia and quadriplegia, or between asymmetri- systematic, reliable, and validated classification system. One
cal hemisyndromes (with very few signs on the contralateral might question whether, with more accurate ways of catego-
side) and bilateral CP. For clinical practice even the recent rizing children according to topographical features of CP or
elegant work of the Surveillance of Cerebral Palsy in Europe types of motor impairment, these classifications could have
group (SCPE 2000) leaves open to judgement how to distin- more clinical utility for prognostic and planning purposes.
guish with precision unilateral from bilateral CP as well as For this to happen, however, there will need to be clearly for-
how to distinguish diplegia from quadriplegia. mulated descriptions of the clinical characteristics of hemi-
Perhaps even more challenging from a clinical perspective plegia, diplegia, triplegia, and quadriplegia based on
is how to describe the type of motor impairment found in peo- international consensus, with evidence that these clinical
ple with CP. Again, the SCPE group has proposed a hierarchical subgroups can be identified reliably by people with appro-
system for the classification of CP subtypes (SCPE 2000, see priate training and experience. The same argument applies
figure p 821), but to the best of our knowledge (MA Johnson, to the need for the application of the SCPE (2000) distinctions
personal communication) this algorithm has yet to be field- to predominant types of motor impairments, assuming that
tested for its reliability and validity. Of all classification sys- these can be shown to be reliable in practice. The develop-
tems, only functional status as categorized by the GMFCS has ment of a reference and training manual for clinicians by the
been objectively demonstrated to be both reliable and valid SCPE study group is now under way and will contribute in

Table IV: Parameters of motor development for classification of function level (GMFCS) combined with classification of
impairment level (limb distribution)

GMFCS level Limb distribution n Mean observations Limit 95% CI 50% range
per child (GMFM-66) (GMFM-66) (GMFM-66)

I Hemiplegia 86 4.2 +2.3 (ns) 82.693.7

Diplegia 80 3.9 87.1 84.689.3 79.392.3
II Diplegia 51 4.5 67.9 64.970.8 59.975.1
Quadriplegiaa 15 4.2 +0.4 (ns) 61.075.9
III Diplegia 68 4.0 56.0 54.057.9 50.461.5
Quadriplegiaa 51 4.1 3.6 (p<0.06) 46.758.0
IV Diplegia 17 3.8 43.3 41.545.2 43.343.3b
Quadriplegiaa 110 3.9 3.7 (p<0.001) 39.639.6b
V Quadriplegiaa 126 3.8 22.3 20.724.2 16.629.5

, difference in GMFM-66 points between value of reference group (children with diplegia): + is a higher value than reference group; is a
lower value than reference value.
aFor this analysis, data on all children with triplegia and quadriplegia were collapsed into quadriplegia group.
bFigures are rounded to one decimal place. For instance, level IV diplegia 50% range is 43.29 to 43.30. Variance for random effect in the limit

parameter is near 0, meaning that there is no evidence for individual differences in limit among these children.
CI, confidence interval; GMFCS, Gross Motor Function Classification System; GMFM, Gross Motor Function Measure; ns, not significant.

466 Developmental Medicine & Child Neurology 2004, 46: 461467

the coming years to a uniform, reliable, and multidimension- Developmental Medicine No. 87. London: Spastics International
Medical Publications (Mac Keith Press). p 111.
al classification of CP (Krgeloh-Man et al. 2003). Jarvis S, Hey E. (1984) Measuring disability and handicap due to
In recent years several international collaborations in the cerebral palsy. In: Stanley F, Alberman E, editors. The Epidemiology
field of CP (Krgeloh-Mann et al. 1993, Palisano et al. 1997, of the Cerebral Palsies. Clinics in Developmental Medicine No.
SCPE 2000) have brought together the experience and ideas 87. London: Spastics International Medical Publications (Mac
Keith Press). p 3545.
of colleagues from many centres. It will be important for clas- Kennes J, Rosenbaum P, Hanna SE, Walter S, Russell D, Raina P,
sification systems to be translated into the language(s) of Bartlett D, Galuppi B. (2002) Health status of school-aged
each country so that consistent systems are applied interna- children with cerebral palsy: information from a population-
tionally. This has already occurred with systems such as the based sample. Dev Med Child Neurol 44: 240247.
GMFCS and for a number of clinical measures now in use or Krgeloh-Mann I, Hagberg G, Meisner C, Schelp B, Haas G, Eeg-
Olofsson KE, Selbmann HK, Hagberg B, Michaelis R. (1993)
being developed around the world. We believe that the time Bilateral spastic cerebral palsy a comparative study between
is ripe for further refinement of the clinical details of the CP south-west Germany and western Sweden. I: Clinical patterns
syndromes in an effort to increase the clarity and consisten- and disabilities. Dev Med Child Neurol 35: 10371047.
cy with which these children are characterized. Such efforts Krgeloh-Mann I, Petruch UR, Gainsborough M, Cans C. (2003) The
Reference and Training Manual (RTM) of the Surveillance of
would make an enormous contribution to the field, to the Cerebral Palsy in Europe, SCPE a video and text based interactive
benefit of everyone for whom CP is important. CD-ROM. Dev Med Child Neurol 45 (Suppl. 97): 6. (Abstract)
Michaelis R, Edebol-Tysk K. (1989) New aetiopathological and
DOI: 10.1017/S0012162204000763 nosological aspects of cerebral palsy syndromes. Giorn
Neuropsichiatr Et Evol Suppl. 4: 2530.
Accepted for publication 30th January 2004. Mutch L, Alberman E, Hagberg B, Kodama K, Perat MV. (1992)
Cerebral palsy epidemiology: where are we now and where are
Acknowledgements we going? Dev Med Child Neurol 34: 547551.
This work was undertaken while JW Gorter was at CanChild, Palisano R, Rosenbaum P, Walter S, Russell D, Wood E, Galuppi B.
supported by a Post-doctoral Fellowship from ZonMW, the (1997) Development and reliability of a system to classify gross
Netherlands Organisation for Health Research and Development motor function in children with cerebral palsy. Dev Med Child
(grant 014-32-034). PL Rosenbaum holds a Canada Research Chair Neurol 39: 214223.
in Childhood Disability from the Canadian Institutes of Health Palisano RJ, Hanna SE, Rosenbaum PL, Russell DJ, Walter SD,
Research (CIHR). SD Walter holds a Senior Investigator award from Wood EP, Raina PS, Galuppi BE. (2000) Validation of a model of
CIHR. P Raina holds a National Health Scholars award from gross motor function for children with cerebral palsy. Phys Ther
CIHR/NHRDP. The Ontario Motor Growth study was made possible 80: 974985.
by the generous support of the Medical Research Council of Canada Pharoah POD, Cooke T, Johnson MA, King R, Mutch L. (1998)
(now CIHR; grant MT-13476), and the National Centre for Medical Epidemiology of cerebral palsy in England and Scotland 19849.
Rehabilitation Research of the National Institute of Child Health and Arch Dis Child 79: F21F25.
Human Development at NIH (grant R01-HD-34947). We thank Pinto-Martin JA, Riolo S, Cnaan A, Holzman C, Susser MW, Paneth N.
David Cameron for his contributions in data management. A (1995) Cranial ultrasound prediction of disabling and
preliminary report of the findings of this study was made by the first nondisabling cerebral palsy at age two in a low birth weight
author at the European Academy of Childhood Disability in Pisa, population. Pediatrics 95: 249254.
Italy on 24 October 2002. Rosenbaum PL, Walter SD, Hanna SE, Palisano RJ, Russell DJ, Raina P,
Wood E, Bartlett DJ, Galuppi BE. (2002) Prognosis for gross
motor function in cerebral palsy: creation of motor development
References curves. JAMA 288:13571363.
Alberman E. (1984) Describing the cerebral palsies: methods of Russell DJ, Avery LM, Rosenbaum PR, Raina P, Walter SD, Palisano RJ.
classifying and counting. In: Stanley F, Alberman E, editors. The (2000) Improved scaling of the Gross Motor Function Measure
Epidemiology of the Cerebral Palsies. Clinics in Developmental for children with cerebral palsy: evidence of reliability and
Medicine No. 87. London: Spastics International Medical validity. Phys Ther 80: 873885.
Publications (Mac Keith Press). p 2731. Russell DJ, Rosenbaum PL, Avery LM, Lane M. (2002) Gross Motor
Badawi N, Watson L, Petterson B, Blair E, Slee J, Haan E, Stanley F. Function Measure (GMFM-66 and GMFM-88) Users Manual.
(1998) What constitutes cerebral palsy? Dev Med Child Neurol London: Mac Keith Press.
40: 520527. Russell DJ, Rosenbaum PL, Cadman DT, Gowland C, Hardy S, Jarvis
Balf CL, Ingram TTS. (1955) Problems in the classification of S. (1989) The gross motor function measure: a means to evaluate
cerebral palsy in childhood. BMJ 2:163166. the effects of physical therapy. Dev Med Child Neurol 31: 341352.
Bax MCO. (1964) Terminology and classification of cerebral palsy. Surveillance of Cerebral Palsy in Europe (SCPE). (2000) Surveillance
Dev Med Child Neurol 6:295307. of Cerebral Palsy in Europe: a collaboration of cerebral palsy
Beckung E, Hagberg G. (2000) Correlation between ICIDH surveys and registers. Dev Med Child Neurol 42: 816824.
handicap code and Gross Motor Function Classification System in Stanley F, Blair E, Alberman, E. (2000) Cerebral Palsies:
children with cerebral palsy. Dev Med Child Neurol 42: 669673. Epidemiology and Causal Pathways. Clinics in Developmental
Beckung E, Hagberg G. (2002) Neuroimpairments, activity Medicine No. 151. London: Mac Keith Press. p 19.
limitations, and participation restrictions in children with Wood E, Rosenbaum P. (2000) The Gross Motor Function
cerebral palsy. Dev Med Child Neurol 44: 309316. Classification System for Cerebral Palsy: a study of reliability and
Blair E, Stanley F. (1985) Interobserver agreement in the classification stability over time. Dev Med Child Neurol 42: 292296.
of cerebral palsy. Dev Med Child Neurol 27: 615622. World Health Organization. (2001) International Classification of
Blair E, Stanley FJ. (1997) Issues in the classification and epidemiology Functioning, Disability and Health ICF. Geneva: World Health
of cerebral palsy. Mental Retard Dev Disabil Res Rev 3:184193. Organization.
Hagberg G, Hagberg G, Olow I. (1975) The changing panorama of
cerebral palsy in Sweden 19541970. II. Analysis of the various
syndromes. Acta Paediatr Scand 64:193200.
Hagberg B, Hagberg G, Beckung E, Uvebrant P. (2001) Changing List of abbreviations
panorama of cerebral palsy in Sweden. VIII. Prevalence and origin
in the birth year period 199194. Acta Paediatr 90: 271277. GMFCS Gross Motor Function Classification System
Ingram TTS. (1984) A historical view of the definition and GMFM Gross Motor Function Measure
classification of the cerebral palsies. In: Stanley F, Alberman E, SCPE Surveillance of Cerebral Palsy in Europe
editors. The Epidemiology of the Cerebral Palsies. Clinics in

Classification of Cerebral Palsy Jan Willem Gorter et al. 467