The relationship of cerebral palsy subtype and functional motor impairment: a population-based study


  • Acknowledgments
    MIS is grateful for the support of the Montreal Children’s Hospital Foundation during the writing of this manuscript. Reseau de Santé Developpement et Bien Etre (RSDBE) of the Fonds de recherche en santé du Quebec FRSQ and the CP Research Chair of Laval University provided operational funding of Registre de la paralysie cérébrale au Québec (REPACQ). Anna Radzioch created the tables. Alba Rinaldi provided secretarial assistance.

Dr Michael Shevell at Room A-514, Montreal Children’s Hospital, 2300 Tupper, Montreal, Quebec H3H 1P3, Canada.


Aim  Traditionally, cerebral palsy (CP) had been classified according to the distribution and quality of motor impairment. A standardized functional classification of gross motor skills has recently been validated – the Gross Motor Function Classification System (GMFCS). The relationship between the neurological subtype of CP and GMFCS level remains undefined in CP.

Method  The Quebec Cerebral Palsy Registry (Registre de la paralysie cérébrale au Québec [REPACQ]) over a 4-year birth interval (1999–2002 inclusive) identified 301 children with CP. Information on both CP subtype and GMFCS level was available for 243 children (138 males, 105 females) with final data extraction at a mean age of 44 months (SD 14mo, range 24–79mo). Proportions of children with a particular CP subtype at GMFCS levels I to III versus levels IV to V were determined and compared.

Results  CP subtype versus GMFCS levels I to III or IV to V was distributed proportionally as follows: spastic diplegic, 51/52 (98%) versus 1/52 (2%); spastic quadriparetic, 20/85 (24%) versus 65/85 (76%); spastic hemiplegic, 76/77 (99%) versus 1/77 (1%); dyskinetic, 4/16 (25%) versus 12/16 (75%); other (triplegic or ataxic–hypotonic), 10/13 (77%) versus 3/13 (23%). These distributions (proportions) all yielded significant (p<0.001) Pearson χ2 values.

Interpretation Neurological subtype is a powerful predictor of functional status related to ambulation. This has implications for counseling families.

Cerebral palsy (CP) is considered the most common cause of physical disability in childhood.1 Its conceptualization dates from the mid-19th century.2 Objective motor findings and impairments of early onset are its core essential features. It is a ‘symptom complex’ characterized by heterogeneous presentations, evolutions, etiologies, comorbid conditions, functional implications, service needs, and outcomes.3 This heterogeneity, combined with a backdrop of a maturing central nervous system and a child who is naturally changing and evolving dynamically with time, have challenged our efforts to classify CP.4

Not surprisingly, several classification schemes have been put forward and adopted into general practice. Traditionally, and from a classical perspective, medical practitioners have used a scheme that uses the quality and topographical pattern of motor impairment to categorize CP into subtypes (e.g. spastic vs dyskinetic vs ataxic-hypotonic vs mixed; quadriplegic vs hemiplegic vs diplegic vs other).5–7 Based on a standard neurological examination, this scheme provides insight into the localization of the congenital or acquired lesion, its pathogenesis, and etiology.8 This traditional classification system provides no insight into a child’s limitations, severity of impairment, or service needs and lacks psychometric validation; however it does offer a shorthand way for health professionals to describe a child’s impairment.9 For the most part, this subtype assignment remains unchanged over a child’s life. However, for a minority (approximately 25%), subtype change may occur, especially for those with a non-spastic subtype, reflecting the interplay of intrinsic and extrinsic influences.9

With the recent emerging interest in function, participation, and activity limitations in neurodevelopmental disability, other function-based classification schemes have emerged.10 The Gross Motor Function Classification System (GMFCS) addresses gross motor functional capability and seeks to capture meaningful differences between categories of children, especially as they relate to the key skill of individual ambulation.11 Psychometrically robust for reliability and validity, it too functions as an effective shorthand way to capture a child’s skills and needs, allowing for clear and concise communication among health professionals.12 Unlike the traditional subtype classification scheme, it does provide insight into severity and needs, with the bonus of accurate prognosis, when applied early in life, for eventual ambulation status.13 Despite its relatively recent development, the GMFCS has had a remarkably rapid uptake into clinical practice and research around the world.14

These two most widely-used classification schemes in CP are clearly complementary, capturing different aspects of the child with CP. They both provide a wealth of valuable information that by definition is different in content and intent. What has yet to be explored in detail is the interrelationship of these classification approaches. Do they independently vary or do they co-vary? The objective of this study was to use a recently implemented population-based CP registry to explore the concurrent distribution of CP subtypes and GMFCS levels. Such information is of value as we continue to seek consensus about the definition and accurate classification of this neurodevelopmental disability.


Children were recruited through the Quebec Cerebral Palsy Registry (Registre de la paralysie cérébrale au Québec [REPACQ]). This Registry was established in 1998 by a multidisciplinary consortium of clinicians and researchers with an interest in CP. After an initial effort over 4 years with international experts and then existing CP registries (i.e. Western Australia,15 Surveillance of Cerebral Palsy in Europe 16) to reach a consensus on the definition and diagnosis of CP, mechanisms of case ascertainment, data variables to be collected, and mechanisms of data collection, the Registry became operational in 2004 in 6 of 17 geographically defined administrative health and social-service regions (island of Montreal, Quebec City, Laurentians, Lanaudiere, Estrie, Outaouais) of the province of Quebec, representing about half of the province’s population and annual births. Cases were ascertained through systematic surveys of regionalized pediatric rehabilitation and medical service providers. Quebec has a universal third-party (government-funded) programme of medical and rehabilitation service provision available to all provincial residents. Regionalization requires that those children needing provision of rehabilitation services are serviced through local specialized centres responsible for a defined geographical region. Access to these centres requires a precise medical diagnosis (i.e. CP) by a pediatric sub-specialist (i.e. pediatric neurologist) before intake. During the birth interval covered by this report, there were approximately 20 practicing pediatric neurologists in the province. Once children were identified, parents or guardians were approached for consent to participate in the Registry. Once consent was obtained, over 120 variables pertaining to each child was sought through a combination of direct review of maternal and child medical and rehabilitation records and parental (preferably maternal) interview. This data review was conducted by trained local research assistants according to standardized policies and procedures established by the Registry. Data collection was supervised consistently by a single individual (LD) with a subset randomly selected for independent validation of accuracy. Ethical permission for the Registry’s establishment and implementation was obtained at the local host institution (McGill University Health Center) and each participating pediatric rehabilitation centre.

Participants were ascertained only when a child was beyond the age of 2 years and, where possible, confirmed at 5 years of age or beyond. CP was defined in accordance with recent consensus statements as a non-progressive motor impairment of early onset, that is presumably cerebral in origin, which may or may not be associated with developmental delays, cognitive disability, language impairment, epilepsy, sensory (auditory or visual) loss, orthopedic abnormalities, or behavioral difficulties.3,9,17 A recognized motor impairment required objective changes in tone, muscle strength, posture, reflexes, and motor skills on examination for diagnosis. Non-progressive referred to the underlying pathological process and not apparent clinical manifestations with genetic and metabolic disorders considered by Badawi et al. generally excluded from consideration.18 Early onset meant signs and symptoms were evident before 1 year of age. By definition, neuromuscular disorders and myelodysplasias were excluded from diagnosis. As noted above, data were obtained in a systematic fashion by trained research assistants according to established written procedures and training protocols, and a subset was checked for accuracy before inscription. Data were obtained on demographic features, maternal medical and obstetric history, labor and delivery of the affected child, neonatal course, age at onset, diagnostic investigations and their results, type of CP, gross motor functional status, coexisting medical and comorbid conditions, and medical and rehabilitation service provision (both previously and at the time of inscription). Data were entered into a secure, computerized database for subsequent analysis.

For this study, participants were children born from 1999 to 2002 (a 4yr birth cohort). When both classification schemes under study were noted for an individual child in their medical and rehabilitation records, present, the subtype of CP and gross motor functional status were extracted. Subtype of CP was classified according to the quality and topographical pattern of motor impairment last documented by an objective assessment with the following well-used clinical scheme: (1) spastic–quadriplegic (spasticity [symmetrical or asymmetrical] in all four limbs with equivalent or greater spasticity in the upper extremities), spastic–hemiplegic (spasticity restricted to one side of the body with relative variable involvement of the upper and lower extremity); spastic–diplegic (spasticity in the lower extremities far in excess of any discernable in the upper extremities, which or may not be relatively asymmetric); or spastic–other (spasticity in one limb or in all limbs except one); (2) dyskinetic (i.e., athetosis, chorea, or dystonia in the absence of objective weakness or tone changes); (3) ataxic–hypotonic.1,5–9 Those with a mixed profile of dyskinetic and spastic features were counted as dyskinetic. Similarly, the last objective assessment was used to assign functional mobility according to the GMFCS.11 This relatively recently developed measure provides an assessment of a child’s gross motor function and abilities in early life, emphasizing the stratification of function into categories that capture meaningful differences for a child and family from most able (level I) to most limited (level V).12 Particular elements of actual motor function at different ages are used to differentiate children and assign categories of motor function. Content validity, reliability, consistency, and predictive validity have been established for this now widely-used measure.19–21 Independent ambulation, without (levels I and II) and with (level III) assistance, is distinguished from non-ambulation (levels IV and V).

All children in the registry for the 4-year birth cohort for whom data on CP subtype and GMFCS level were available were included for data analysis. The distribution of CP subtype and GMFCS level were determined both uniquely and concurrently. The distribution of proportions of GMFCS level (dichotomously grouped into levels I to III [i.e. ambulatory] and levels IV and V [non-ambulatory]) within a CP subtype was compared by Pearson χ2 analysis with a p value of 0.05 or less (two-tailed) established a priori as the threshold for statistical significance.


In all, 301 children with CP were identified in the six administrative regions comprising REPACQ over the 4-year birth interval (1999–2002 inclusive). Given approximately 144 000 live births in these same regions over this interval, this yields a crude prevalence of 2.09 per 1000 live births. Concurrent information on CP subtype and GMFCS level was available for 243 children, which comprises the study cohort (for the remaining 58 children, parental/guardian consent could not be obtained for chart review and interview). Participants included 138 males and 105 females. The mean age of last data extraction for Registry inscription was 44 months (SD 14mo, range of 24–79mo). Distribution of documented CP subtypes is noted in Table I. Eighty-nine per cent (218/243) had spastic CP, 35% (85/243) quadriplegic, 31% (77/243) hemiplegic, 21% (52/243) diplegic, 2% (4/243) other (triplegic or monoplegic), 7% (16/243) dyskinetic CP, and 4% (9/243) ataxic–hypotonic CP. Distribution of GMFCS levels is highlighted in Table II. Grouping together dichotomously levels I to III and levels IV and V, 66% (161/243) were at levels I to III (i.e. capable of independent ambulation [assisted or non-assisted]) and 34% (82/243) at levels IV and V (i.e. incapable of independent ambulation).

Table I.   Distribution of cerebral palsy subtypes in study participants
Subtypen (%)
Spastic quadriplegia85 (35)
Spastic hemiplegia77 (31)
Spastic diplegia52 (21)
Dyskinetic16 (7)
Ataxic–hypotonic9 (4)
Other4 (2)
Table II.   Distribution of GMFCS levels in study participants
GMFCS leveln (%) 
  1. GMFCS, Gross Motor Function Classification System.

I108 (44)66% (I–III)
II23 (10)
III30 (12)
IV43 (18)34% (IV–V)
V39 (16)

The concurrent distribution of CP subtypes and GMFCS levels is shown in Table III. Seventy-six per cent (65/85) of children with spastic quadriplegia were at levels IV or V. Ninety-nine per cent (78/79) of children with spastic hemiplegia were at levels I to III, whereas similarly 98% (51/52) children with spastic diplegia were at levels I to III. Of the children with dyskinetic CP, 75% (12/16) were at levels IV or V. From the perspective of GMFCS level, 93% (40/43) of those at level IV had either spastic quadriplegic or dyskinetic CP and 95% (37/39) of those at level V had the same subtype attribution to either spastic quadriplegia or dyskinetic.

Table III.   Distribution of GMFCS level by neurological subtype
Neurological subtypeGMFCS level
  1. GMFCS, Gross Motor Function Classification System.

Spastic quadriplegia36113332
Spastic hemiplegia686201
Spastic diplegia3171310

As summarized in Table IV, Pearson χ2 analysis of proportions of CP subtype falling into either levels I to III or levels IV and V revealed that the distributions for each subtype compared with all other subtypes invariably reached statistical significance. Similarly, and not unexpectedly, grouping together spastic quadriplegic and dyskinetic subtypes also yielded a statistically significant Pearson χ2 analysis. Sex did not influence these proportional distributions.

Table IV.   Pearson χ2 test, neurological subtypes of cerebral palsy (CP) × GMFCS level (I–III vs IV–V)
Neurological subtypeGMFCS levelχ2p value
  1. Degrees of freedom=1. GMFCS, Gross Motor Function Classification System.

Spastic quadriplegia2065106.7<0.001
Other CP14117
Spastic hemiplegia176153.1<0.001
Other CP8581
Spastic diplegia51130.0<0.001
Other CP11081
Other CP15770
Spastic quadriplegia or dyskinetic2477139.6<0.001
Other CP1375


The population-based registry used for inclusion of participants in this study yielded a crude prevalence ratio of 2.09 cases of CP per 1000 live births. This is in the middle of the reported range for the prevalence for CP (1.5–2.5 per 1000 live births) in developed regions comparable to Quebec.22 This result suggests reasonably complete case ascertainment. Similarly, the profile of CP subtype distribution23 and GMFCS level attribution is also quite similar to that noted in the recent literature.24 These observations suggest that a relatively unbiased sample of children with CP was studied.

Our data clearly demonstrate a tight link between CP subtype and GMFCS level across all subtypes that occur with any appreciable frequency (the low number of children with the ataxic–hypotonic variant effectively precludes conclusive analysis and statements on this subtype). This is especially so for the key aspect of gross motor function, which is the ability, eventually, to ambulate independently. It is highly unusual, and indeed a rare event (2/129 [1.5%]), that a child with either spastic diplegia or spastic hemiplegia does not achieve independent ambulation. This is a key parental question at the time of initial diagnosis of their child. Independent ambulation is also a major determinant of participation that prefaces an improved individual quality of life, however defined.25,26 For the child with either spastic quadriplegia or dyskinesia, independent ambulation occurs only in a minority (24/101 [23.7%]). Thus it can be emphatically asserted that neurological subtype, once known, is a powerful predictor of functional status related to ambulation. Conversely, and not unexpectedly, assignment of functional status provides insight into what neurological subtypes can be expected. From our data, 95% (77/82) of children at GMFCS levels IV or V (i.e. those who are non-ambulatory) had either the spastic quadriplegic or dyskinetic subtypes.

Although these two classification schemes are clearly complementary, they are relationally informative to a high degree. This has clear clinical and taxonomic implications. Determining a cerebral subtype goes beyond the envisioned descriptive emphasis of those who originally formulated this approach.5–7 It has already been shown that it has etiological implications.8 It also appears to have implications for the burdens of comorbidity.27 To this, it can be added that gross motor function is highly correlated to a child’s CP subtype. Thus, beyond its descriptive qualities, a CP subtype imparts a profile of possible expectations that have a discernible impact for the initial etiological evaluation, ongoing programmatic approach, and expectations for a particular child. Informing expectations based on the initial, and often relatively early, recognition of a subtype, provides an easy way for the clinician to inform and guide prognosis, counseling, and systematic evaluation imperatives, both immediately and as the child grows. What remains to be determined is what factors (if any), intrinsic or extrinsic to the child and family, determine the attainment of levels I to III or IV and V among those children with either a spastic quadriplegic or dyskinetic subtype variant.

Our findings also have some taxonomic implications. The strong relationship between a neurologically based subtype and gross motor function renders more robust, and perhaps more conceptually valid, the information imparted by both classification approaches.2 This is especially important for classification of neurological subtype as it has evolved historically and has not been subject to the methodological rigors undertaken in the development of the GMFCS.12 Clearly the complementary approaches satisfy differing information needs and challenges.28 The differing approaches are most certainly not redundant nor exclusionary, and they capture alternative aspects of the same entity. A multimodal approach to conceptualizing, through varying classification schemes, the child with CP, offers the best hope of accurately capturing issues related to pathogenesis, current status, needs, and future expectations.9 It also can best shape future research efforts directed at homogeneous subgroups and delineating change as a result of specific interventions. A major challenge, already alluded to by others, will be incorporating additional information about comorbidities and pathogenesis into classification schemes.3,4 Whether classified separately or within, it remains to be determined if such information is either necessary or enlightening, or whether it merely makes matters more complex and cumbersome for the practice situation.

Our data suggest that there is a very strong relationship between neurological subtype and GMFCS level, especially stratified dichotomously for independent ambulation. Accurate determination in the clinic of neurological subtype remains an important element of the evaluation of the child with CP. Although not invariable, a profile for the child can be created based on this simple, routine evaluation, which can then inform and shape a future path at a variety of levels: individual, familial, medical, and rehabilitation. An accurate, well-informed path is a necessary precondition to achieving the optimal outcome that facilitates broad participation for the affected child.

Members of the Registre de la paralysie cérébrale au Québec (REPACQ) consortium

Marie-Danielle Boucher MD (Quebec City), Claude Desjardins MD (Hull), Josée Fortier MD (Trois-Rivières), Louise Koclas MD (Montreal), Céline Lamarre MD (Montreal), Francine Malouin PT PhD (Quebec City), Jean Mathieu MD MSc (Chicoutimi), Diane Munz MD (Montreal), Nicole Pigeon MD (Sherbrooke), and Carol L Richards PT PhD (Quebec City).