You are on page 1of 8

DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY

ORIGINAL ARTICLE

Stability and decline in gross motor function among children and youth with cerebral palsy aged 2 to 21 years
STEVEN E HANNA

| PETER L ROSENBAUM MD FRCP(C) 2 | DOREEN J BARTLETT PT PHD 3 | ROBERT J PALISANO PT SCD 4 | STEPHEN D WALTER PHD 1 | LISA AVERY MSC 5 | DIANNE J RUSSELL PHD 6
PHD

1 Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada. 2 Faculty of Health Science, McMaster University, Hamilton, Ontario, Canada. 3 School of Physical Therapy, University of Western Ontario, London, Ontario, Canada. 4 Programs in Physical Therapy and Rehabilitation Science, Drexel University, Philadelphia, PA, USA. 5 ASQME Study of Quality of Life, Mobility and Exercise, CanChild Centre for Childhood Disability Research, McMaster University, Hamilton, Ontario, Canada. 6 School of Rehabilitation Science, McMaster University, Hamilton, Ontario, Canada. Correspondence to Dr Steven E Hanna at CanChild Centre for Childhood Disability Research, McMaster University, IAHS 411, 1400 Main Street West, Hamilton, Ontario L8S 1C7, Canada. E-mail: hannas@mcmaster.ca

PUBLICATION DATA

Accepted for publication 7th September 2008. Published online


ACKNOWLEDGMENTS

This research was supported by a grant for the ASQME project from the Canadian Institutes of Health Research (CIHR; MOP-53258). The OMG study was supported by grants from the CIHR (MT-13476) and the National Center for Medical Rehabilitation Research of the US National Institute of Child Health and Human Development (RO1-HD-34947). Dr Rosenbaum holds a CIHR Canada Research Chair in Childhood Disability and the Scotiabank Chair at the McMaster Child Health Research Institute. Dr Russell holds a Research Scholar award from the Ontario Federation of Cerebral Palsy. The CanChild Centre for Childhood Disability Research is supported by the Ontario Ministry of Health and Long-term Care. We thank the team of ASQME assessors across Ontario who worked to collect these data. Special thanks are owed to Barb Galuppi who coordinated the OMG and ASQME studies and supervised the collection of these data. We are indebted to the children, adolescents, and their families who were involved in these studies and were so willing to share their experiences.

This paper reports the construction of gross motor development curves for children and youth with cerebral palsy (CP) in order to assess whether function is lost during adolescence. We followed children previously enrolled in a prospective longitudinal cohort study for an additional 4 years, as they entered adolescence and young adulthood. The resulting longitudinal dataset comprised 3455 observations of 657 children with CP (369 males, 288 females), assessed up to 10 times, at ages ranging from 16 months to 21 years. Motor function was assessed using the 66-item Gross Motor Function Measure (GMFM-66). Participants were classied using the Gross Motor Function Classication System (GMFCS). We assessed the loss of function in adolescence by contrasting a model of function that assumes no loss with a model that allows for a peak and subsequent decline. We found no evidence of functional decline, on average, for children in GMFCS Levels I and II. However, in Levels III, IV, and V, average GMFM-66 was estimated to peak at ages 7 years 11 months, 6 years 11 months, and 6 years 11 months respectively, before declining by 4.7, 7.8, and 6.4 GMFM-66 points, in Levels III, IV, and V respectively, as these adolescents became young adults. We show that these declines are clinically signicant.

Cerebral palsy (CP) is a group of disorders affecting development of movement and posture, causing activity limitations attributed to non-progressive disturbances that occur in the developing fetal or infant brain.1 Although non-progression of the underlying neuropathology is a dening feature of the disorder, this does not apply to the clinical manifestations, which are widely believed to change through the lifespan.2 In particular, after developmental

gains in childhood,3 persons with CP may decline in gross motor function as they move into adolescence and young adulthood.4 Some studies have documented a reduction in the number of people with CP who walk after age 18.57 The timing, scope, and extent of change in gross motor function are not well documented. Previous research to document declines in gross motor function with age has involved self-reports or chart reviews
The Authors. Journal compilation Mac Keith Press 2009 DOI: 10.1111/j.1469-8749.2008.03196.x 1

of changes in ambulation. Bottos et al.5 conducted a chart review and found that 13 of 29 persons with CP who could walk independently before age 18 years lost this capability between the ages of 20 and 40 years. In a survey of 766 Norwegian adults with CP, of the 88% who had previously walked to some degree, 45% reported that their walking ability had deteriorated, whereas 27% said it had improved.6 The largest studies of change in ambulation come from longitudinal data collected by the California Department of Developmental Services.7,8 The Department collects annual assessments across a variety of health and function domains from all persons who receive disability-related services in the state. Day et al.7 conducted a secondary analysis of the resulting data for 7550 children with CP who were rst assessed at ages 9 to 12 years, and who were followed for 15 years. They constructed an ordinal assessment of usual mobility, ranging from (1) walks and climbs well to (4) does not walk. Children who walked unsteadily and sometimes used a wheelchair were most likely to lose ambulatory ability as young adults (34%). Thus, it seems likely that the probability of decrease in mobility depends on the initial degree of mobility limitation. In contrast to studies that focused on changes in ambulation and wheelchair use, Rosenbaum et al.3 measured gross motor function comprehensively, systematically, and repeatedly, using the standardized and validated Gross Motor Function Measure (GMFM-66) in a large, representative, prospective longitudinal cohort of children with CP who were 12 years of age and younger at enrolment to the study. In this study (the Ontario Motor Growth [OMG] study), nonlinear models of motor development were estimated assuming that childrens GMFM-66 scores rise toward a limit of motor ability, which varies according to their Gross Motor Function Classication System (GMFCS) level.9 This model worked well for characterizing early developmental gains. However, in adolescence and young adulthood, the level of motor performance achieved earlier may not be maintained. We recently completed a study that followed a sample of the participants in the OMG3 study into adolescence and young adulthood, with continuing longitudinal assessments using GMFCS and GMFM-66. In the current report, we aim to assess the possible loss of gross motor function during adolescence. We do this by contrasting a model of motor function that assumes no loss of function with a model that allows for a peak and subsequent decline. We identify the GMFCS levels in which loss of gross motor function is most likely, and estimate the average age of onset and amount of lost function.
2 Developmental Medicine & Child Neurology 2009

METHOD Participants The data for this study were obtained as part of the Adolescent Study of Quality of life, Mobility and Exercise (ASQME). ASQME is a 5-year continuation of the OMG study, which followed a stratied random sample of 657 children with CP from a population-based cohort obtained between 1996 and 2001.3 Sampling and procedures for the OMG study are described in detail elsewhere.3 Both studies were approved by the Research Ethics Board at McMaster University and several community programs. A parent provided informed consent and each adolescent provided informed consent or assent, if possible. Children with CP and their families were recruited to the OMG study through 19 regional ambulatory childrens rehabilitation centers in the province of Ontario, Canada, beginning in 1996. These centers are publicly funded and each program serves the majority of eligible children with CP in its geographical area. The OMG sample was stratied by GMFCS severity level9 and age. Children completed up to seven assessments during the OMG study (mean=4.0); the childrens ages ranged from 1 year 5 months to 15 years, although 90% of the assessments occurred when children were between 3 years 1 month and 13 years of age. At the completion of that study, all 343 children aged 11 years or older were invited to participate in ASQME. Two hundred and forty-four (71.1%) children and families agreed to participate and 229 (66.8%) completed at least one of four scheduled annual assessments (mean=3.6). For the current report, the OMG and ASQME data were combined to produce a single longitudinal dataset comprising 3455 observations of 657 children with CP, assessed at ages ranging from 16 months to 21 years. The median number of assessments per child in the combined dataset was ve, with 9.0% of children completing one or two assessments, 60.7% completing three to six assessments, and 30.3% completing seven to ten assessments. The median age for an assessment was 9 years 5 months, with 50% of assessments done between 6 and 13 years 2 months of age. Males (n=369) comprised 56% of the sample. Of the 640 children for whom topographical distribution was recorded, 98 (15%) had hemiplegia, 217 (34%) had diplegia, 62 (10%) had triplegia, and 268 (41%) had quadriplegia. In the 639 children for whom type of motor impairment was recorded, 500 (78%) spastic, 39 (6%) dyskinetic, 16 (2.5%) ataxic, 26 (4%) hypotonic, and 58 (9%) mixed types were observed. Measures and Procedure As described by Rosenbaum et al.,3 children in the OMG study who were under 6 years of age were assessed every 6 months, whereas older children were assessed every 9 to 12

months. The 831 assessments of the 229 adolescents in the ASQME study were made annually. Assessments were completed by physical therapists who were trained to administer the measures and were reassessed annually. They classied children according to the GMFCS9 and administered the GMFM-66.10

Gross motor function classification system The GMFCS is a ve-level system that classies children with CP in terms of functional ability and limitations.9,11 The GMFCS is straightforward, valid, and reliable for children under 12 years of age.12 Children in Level I walk without limitations in all settings, whereas children in Level V have severe limitations in head and trunk control, and in self-mobility. Although the GMFCS was recorded at each observation, we followed the OMG study3 by using the rst available GMFCS classication to stratify analyses. This ensured consistency with our previous research. Gross motor function measure Gross motor function was measured using the 66-item GMFM-66, an evaluative instrument for assessing children with CP.10,1315 The GMFM-66 measures capability, or what a child can do in a standardized environment. Items include tasks related to lying and rolling, sitting, crawling and kneeling, standing, walking, running and jumping, with the most difcult items on the scale representing abilities obtained by children developing typically by 5 years of age. Each item is scored by observation on a four-point ordinal scale (03). The GMFM-66 was developed using Rasch analysis;14,15 the underlying Rasch model estimates the difculties of the items,10 so that a childs total ability score can be easily related to the probability of attaining common motor milestones. Total scores can range from 0 to 100. Data analysis Data for each of the ve GMFCS levels were analyzed separately. Within GMFCS level, total GMFM-66 scores were analyzed as nonlinear functions of the childrens ages. To evaluate whether the average pattern of change over age included a peak and decline in function, we tted and contrasted two nonlinear models for each GMFCS level. The rst model (Stable Limit model) is the same as that used in the OMG study3 to t motor development curves from ages 2 to 12 years. Children are assumed to start with GMFM-66 scores near 0 as newborn infants, and to acquire gross motor skills rapidly while they are young, with the rate of change slowing as they reach the limit of their potential. The OMG study3 showed that this rate and limit vary considerably among children with CP,

even within GMFCS level. The Stable Limit model has two parameters, corresponding to the rate and limit of motor change as a function of age. The results of the Stable Limit model are contrasted with a Peak and Decline model. Within the age range 2 to 21 years, the Peak and Decline model will make predictions very similar to the Stable Limit model if there is no evidence of decline in GMFM-66 scores. However, the Peak and Decline model has one extra parameter, and this allows for predictions that may show a peak and decline in score, before a long-term limit. Mathematical details of the two models are given in Appendix SI (Supporting Information published online). To account for the longitudinal structure of the data, both models were estimated as nonlinear mixed-effects models.16,17 The mixed-effects framework allows for the estimation of the average pattern of change within GMFCS level. In addition, orderly variations in the pattern of development between children are accommodated by allowing for individual variation in the parameters that characterize the change function. In effect, individual motor development curves are tted for each child. The nonlinear mixed-effects models were estimated by maximum likelihood using the NLME software routines implemented in the S-Plus statistical programming language.18 To evaluate whether a peak and decline in GMFM-66 occurs, we compared the overall t of the models using the Akaike Information Criterion (AIC) statistic, which is a standard measure of model t.19 Models with lower AIC t better to the data. AIC incorporates a penalty for models having more parameters, so if there is no peak and decline evident in the data, the simpler Stable Limit model will be preferred to the Peak and Decline model. For GMFCS levels where a peak and decline is discovered, we used the estimated parameters of the Peak and Decline model to estimate the average age at peak GMFM-66, and we calculated the average size of the decline from peak GMFM-66 to the estimated long-term limit of GMFM-66. Further detail is given in Appendix SI.

RESULTS Table I shows mean GMFM-66 by GMFCS and age. Although this simple summary uses crude age bands, and does not account for the multiple observations per child, it suggests that the nonlinear trend in GMFM-66 over age may differ by GMFCS level. The trend in means shows no evidence of peak and decline for Levels I and II, whereas there is a suggestion of such a decline for levels IV and V. The trend for level III is not clear. The results of tting the nonlinear mixed-effects Stable Limit and Peak and Decline models are given in Table II. The two parameters for the Stable Limit models are
Motor Function Curves in CP Steven E Hanna et al. 3

Table I: Mean Gross Motor Function Measure (GMFM-66) by age and Gross Motor Function Classification System (GMFCS)
GMFCS 26 Level I (na=183) Mean 95% CI SD nb Level II (na=80) Mean 95% CI SD nb Level III (na=122) Mean 95% CI SD nb Level IV (n =137)
a

Age (y) 69 84.0 82.8, 85.2 8.7 201 60.1 63.2, 66.9 8.6 82 53.0 52.0, 54.3 6.9 131 41.4 40.2, 42.6 7.2 146 24.8 23.5, 26.0 8.4 171 912 86.9 85.5, 88.4 10.0 188 71.0 69.1, 72.8 9.1 97 52.8 51.7, 54.9 7.2 153 39.2 38.0, 40.5 7.8 154 21.0 19.3, 22.6 9.6 133 1216 87.1 85.5, 88.7 10.7 177 70.6 68.7, 72.4 7.2 63 50.8 49.0, 52.7 10.0 114 38.6 36.1, 40.3 9.0 108 21.9 20.1, 23.7 8.7 96 1621 87.6 85.9, 89.3 11.2 166 70.8 69.1, 72.5 7.8 83 51.7 50.0, 53.4 10.0 132 34.4 32.7, 36.2 9.5 115 22.7 19.8, 25.6 12.5 75

72.5 71.1, 73.9 11.1 244 55.7 54.3, 57.1 8.1 139 49.7 48.7, 50.8 6.6 150 38.8 37.7, 40.0 7.6 174 23.1 21.9, 24.2 7.6 162

Mean 95% CI SD nb

Level V (na=135)

Mean 95% CI SD nb

The number of children. bThe number of observations, which typically includes multiple observations per child, within and across age

categories. CI, condence interval.

estimates of the rates and limits of motor change. Consistent with our previous research,3 the clinical interpretability of the rate parameter has been improved by transforming it to Age90. This is easily interpreted as the average age at which children achieve 90% of their expected limit in GMFM-66 motor ability. Thus, smaller values of Age90 (in years) indicate faster progress toward motor development limits. The Peak and Decline models are governed by three parameters, including a limit that is interpreted as in the Stable Limit models. For GMFCS Levels I and II, the Stable Limit and Peak and Decline models produce similar limit estimates, whereas the Peak and Decline limits are lower than those for the Stable Limit model for Levels III to V. The other parameters in the Peak and Decline model are the predicted GMFM-66 at 6 years of age, and a rate parameter. The Peak and Decline rate parameter cannot be transformed to a clinically meaningful quantity, such as Age90, so its interpretability is limited. Instead, we use all the Peak and Decline model parameters to calculate the predicted age at which peak GMFM-66 score is achieved for the average child. Based on comparisons of AICs, the Stable Limit model ts better for GMFCS Levels I and II, whereas the Peak and Decline model ts better for Levels III to V.
4 Developmental Medicine & Child Neurology 2009

This suggests that the average patterns of GMFM-66 include peaks and declines before arriving at the limits of gross motor ability for Levels III to V, but not Levels I and II. Examination of the residual standard deviations (SDs) for each group supports the model choices based on AIC. The SDs indicate the average degree of withinchild deviations of observed GMFM-66 scores from childrens individualized trajectories predicted by the Stable Limit and Peak and Decline models. The SDs for the Stable Limit and Peak and Decline models are comparable for GMFCS Levels I and II, reecting the similarity of predictions when no peak and decline is present. However, the residual SDs in Table II are considerably smaller for the Peak and Decline than Stable Limit models for GMFCS III to V, reecting the predictive advantage for a model that can accommodate peak and decline in GMFM-66. Inspection of the standardized residuals for the best tting model in each level suggested that they conform to the assumptions of the nonlinear mixedeffects model. The predicted average GMFM-66 trajectories from the best models in each GMFCS level are illustrated in Figure 1. The dashed lines illustrate the timing of the peak and highlight the loss from the peak to age 21 years.

Table II: Parameters of motor development for the Stable Limit (SL) and Peak Decline (PD) models by Gross Motor Function Classification System (GMFCS)
GMFCS Level I (n=183) Mean number of observations per child AIC model t SL model SL model PD model GMFM-66 limit (95% CI) Age90 (95% CI) Residual SD PD model GMFM-66 limit (95% CI) Predicted GMFM-66 at age 6y (95% CI) Rate parameter (95% CI) Residual SD
a

II (n=80) 5.8

III (n=122) 5.2

IV (n=137) 5.1

V (n=135) 4.7

5.3

6091a 6164 89.5 (87.7, 91.0) 5y 2mo (4y 10mo, 5y 8mo) 3.9 87.1 (85.4, 88.9) 78.7 (77.1, 80.3) 0.959 (0.952, 0.966) 3.6

2719a 2821 68.5 (66.0, 71.0) 4y 11mo (4y 4mo, 5y 6mo) 3.0 67.2 (64.9, 69.3) 60.8 (58.4, 63.2) 0.949 (0.937, 0.961) 3.0

4026 3848a 53.5 (51.9, 55.0) 3y 2mo (2y 8mo, 3y) 3.4 47.4 (44.2, 50.6) 51.2 (49.9, 52.5) 0.977 (0.974, 0.980) 2.4

4165 3930a 39.5 (38.2, 40.9) 3y 2mo (2y 8mo, 3y 1mo) 3.6 33.5 (30.7, 36.3) 41.1 (39.7, 42.5) 0.979 (0.977, 0.981) 2.5

3984 3820a 22.1 (20.6, 23.6) 2y 5mo (1y 6mo, 3y 8mo) 4.0 17.3 (14.0, 20.6) 23.6 (21.9, 25.2) 0.981 (0.978, 0.984) 3.0

The better tting model (SL or PD). AIC, Akaike Information Criterion; GMFM-66, Gross Motor Function Measure; CI, condence interval.

10 20 30 40 50 60 70 80 90 100

Level I

Level II

Level III* Level IV* Level V*

2 3 4 5 6 7 8 9

11 13 15 17 19 21 Age, yrs

Figure 1: Predicted Gross Motor Function Measure (GMFM-66) motor scores as a function of age by Gross Motor Function Classification level. *GMFCS levels with significant average peak and decline. Dashed lines illustrate age and score at peak GMFM-66.

Table III provides the estimated parameters of the best tting models, the estimated average age of the peak GMFM-66, and average decline from peak to limit, for

GMFCS Levels III to V. The estimated peaks occur well before adolescence (7y 11mo, 6y 11mo, and 6y 11mo respectively). The predicted declines in GMFM-66 range from 4.7 points in GMFCS Level III to 7.8 points in Level IV. To illustrate the clinical consequences of these declines, Table IV relates scores at the average peak and decline to expected performance on selected GMFM items. The probabilities of completing these items (i.e. scoring the maximum, 3) were computed from established Rasch model results for the GMFM-66.10 Table IV shows that the estimated declines can have a clinically meaningful effect on the probability of performing items that are most diagnostic of motor ability in each level. These are items with moderate probabilities at peak function, which are highlighted in Table IV. For example, with an average peak score of 52.2, children in GMFCS Level III have moderate probabilities of successfully completing the standing and supported walking items 35, 68, and 53. The relatively small predicted loss of 4.7 points from peak to limit dramatically decreases the probability of success for these items. The larger predicted loss of 7.8 points for children in GMFCS Level IV all but eliminates the moderate probabilities of attaining the sitting and four-point mobility items 34, 41, and 44.
Motor Function Curves in CP Steven E Hanna et al. 5

GMFM-66

Of course, not all children will experience the average pattern of change for their level. The nonlinear mixedeffects model estimates individual differences in all parameters of change as variance components. These are reported in Table III for each of the parameters, as well as

for the resulting peak and decline estimates, transformed to 50% ranges for ease of interpretation. The 50% ranges give the values of the parameter expected for the middle 50% of children. Note that they are not condence intervals: rather, they give an indication of the expected variabil-

Table III: Peak Decline estimates and between-child variation in the parameters of change for best fitting models, by Gross Motor Function Classification System (GMFCS)a
GMFCS Stable Limit models I Average GMFM-66 limit 50% range Average Age90 50% range Average predicted GMFM-66 at age 6y 50% range Peak Decline Age at peak (y) (95% CI) 50% range GMFM-66 loss from peak to limit (95% CI) 50% rangeb
a

Peak Decline models III 47.4 38.7, 56.1 IV 33.5 26.1, 41.0 V 17.3 9.7, 25.0

II 68.5 61.0, 75.2 4y 11mo

89.5 81.0, 94.4 5y 2mo

4y 1mo, 6y 8mo 3y 7mo, 6y 6mo 51.2 47.1, 55.2 7y 11mo (6y 10mo, 9y) 4.7 (1.9, 7.5) 0, 13.1 41.1 36.2, 46.0 6y 11mo 23.6 17.6, 29.6 6y 11mo 3y 6mo, 10y 4mo 6.4 (2.8, 10.0) 0, 16.6

(6y 2mo, 7y 6mo) (5y 10mo, 7y 11mo) 7.8 (4.9, 10.6) 0, 16.9

3y 8mo, 12y 1mo 4y 6mo, 9y 2mo

Stable Limit models are used for Levels I to II, and Peak and Decline models are used for Levels III to V. Not all parameters are estimated

for both models. Fifty per cent ranges are not reported for the Peak and Decline rate parameter because it has poor interpretability and low between-child variation. bNegative peak loss estimates are undened for the Peak and Decline model, and so the lower bound of the 50% range has been set to 0. GMFM-66, Gross Motor Function Measure; CI, condence interval.

Table IV: Probabilities of attainment (scoring 3) for selected Gross Motor Function Measure items, at the average peak and limit score for Gross Motor Function Classification Levels III to Va
Item Level III Peak (52.2) 10. Prone, lifts head 23. Sitting, arms propping, 5s 22. Supported at thorax, lifts head to midline, 10s 34. Sits on bench, 10s 41. Prone, attains four-point 44. Four-point crawl or hitch, 10f (3.048 m) 35. From standing, sits on bench 68. Walks 10 steps, one hand held 53. Stand arms-free, 3s 69. Walks 10 steps, no support
a

Level IV Peak (41.3) 96.6 97.3 95.6 59.3 48.2 31.2 3.3 1.4 2.3 <1 Limit (33.5) 88.0 89.5 84.0 10.0 3.6 1.0 <1 <1 <1 <1

Level V Peak (23.7) 52.5 37.8 38.4 <1 <1 <1 <1 <1 <1 <1 Limit (17.3) 17.9 4.0 9.1 <1 <1 <1 <1 <1 <1 <1

Limit (47.5) 98.8 99.1 98.5 85.7 85.7 87.0 35.3 19.0 16.1 1.6

99.5 99.6 99.3 93.6 94.6 97.0 74.8 57.7 40.1 14.6

Items with moderate probabilities (2575%) at peak predicted score are in bold.

6 Developmental Medicine & Child Neurology 2009

ity in the change pattern among children. Although the average patterns of change in Levels III to V show a decline in GMFM-66, the 50% ranges indicate that the degree of this decline varies considerably among children.

DISCUSSION Williams20 observed that children with disability grow into adults with a disability, and has called for more attention to adolescence as a key time of transition and an understudied area of disability research and service. This is particularly relevant for youth with CP because therapeutic and support services are also in transition during this period; care often shifts from childrens treatment centers, school support, and early childhood education programs to less intensive and less available adult primary care and rehabilitation. The present study complements and sharpens the suggestive ndings of Day et al.7 Previous work has focused on ambulation, and thus on the higher-functioning children and youth. By contrast, our study presents a detailed and comprehensive account of changes in adolescent gross motor function across the entire spectrum of CP. Previous research suggested that the likelihood of loss might be related to initial overall activity limitation, but none used a validated and reliable classication of activity limitations.7 By using the ve levels of the GMFCS, it is possible to show in detail where functional peak and loss are likely to happen and when in development they occur. This information is now specic enough to be of practical use to clinicians, children, and families. The ndings indicate that children and youth in Levels III, IV, and V are at risk of losing motor function, with the greatest declines apparent in Level IV. Although the specic explanation for these ndings requires further prospective study, we can speculate that some combination of physical growth and decreased self-initiated motor function (perhaps related to the effort involved) may be associated with a natural tendency toward increased energy costs, contractures, and muscle stiffness that have been somewhat held at bay in the earlier years. Changes in spinal alignment may also be relevant. We are currently exploring this question using additional clinical data collected systematically in the ASQME study. Consistent with these ideas, Bottos and Gericke21 and Pimm22 suggest that, among adolescents and young adults, the transition to adult body size and the increasing imbalance between physiological resources and changing environmental demands over time may mean that the maintenance of previously achieved motor function becomes prohibitively demanding for some persons. Jahnsen et al.23 found that fatigue was higher among adults with CP than in the general population. Murphy et al.24 found that joint deterioration or pain was a common rea-

son for adults to stop walking independently later in life, but not before age 25. Because we did not rely on self-reports of usual mobility, it is possible to distinguish what children can do (capacity) from what they choose to do (performance). Our use of a standardized measure suggests it is more likely that the peak loss observed here is related to childrens capacity, rather than their preferences. We exploited the Rasch scaling of the GMFM-66 to show that the average losses of 4.7 to 7.8 points are large enough to produce clinically meaningful changes in the performance of some important gross motor tasks (see Table IV). The implications for usual mobility in diverse settings remain to be seen. Our longitudinal study design allows us to comment on between-child variations in the changes in function. This is important because of the large variability within levels. Clinicians and families should not assume that the average pattern applies to all children in a GMFCS level. The substantial variation in the degree of loss estimated from these data emphasizes that children in Levels III to V are not destined to lose function in adolescence. Rather, we estimate the risk of loss and its variations. The challenges now are to predict more accurately who will experience peak loss within levels and how it can be mitigated. For example, Damiano4 and others have suggested that exercise, physical activity, and participation are crucial to maintaining function as persons with CP age. Obesity may also be an important factor, and we are conducting further work to investigate this. We believe that our observations about peak and loss of gross motor function will be useful for service providers and families, and for researchers seeking to describe the life course of young people with CP.

SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article: Appendix SI: Statistical Appendix (The Stable Limit Model and the Peak Decline Model) Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. REFERENCES
1. Rosenbaum P, Paneth N, Leviton A, et al. A report: the denition and classication of cerebral palsy April 2006. Dev Med Child Neurol 2007; 49: (Suppl. 109) 814. Erratum in: Dev Med Child Neurol 2007; 49: 480. 2. Dan B. Progressive course in cerebral palsy? Dev Med Child Neurol 2007; 49: 644.

Motor Function Curves in CP Steven E Hanna et al. 7

3. Rosenbaum PL, Walter SD, Hanna SE, et al. Prognosis for gross motor function in cerebral palsy: creation of motor development curves. J Am Med Assoc 2002; 288: 135763. 4. Damiano D. Activity, activity, activity: rethinking our physical therapy approach to cerebral palsy. Phys Ther 2006; 86: 153440. 5. Bottos M, Feliciangeli A, Sciuto L, Gericke C, Vianello A. Functional status of adults with cerebral palsy and implications for treatment of children. Dev Med Child Neurol 2001; 43: 51628. 6. Jahnsen R, Villien L, Egeland T, Stanghelle JK, Holm I. Locomotion skills in adults with cerebral palsy. Clin Rehabil 2004; 18: 30916. 7. Day SM, Yvonne WW, Strauss DJ, Shavelle RM, Reynolds RJ. Change in ambulatory ability of adolescents and young adults with cerebral palsy. Dev Med Child Neurol 2007; 49: 64753. 8. Strauss D, Ojdana K, Shavelle R, Rosenbloom L. Decline in function and life expectancy of older persons with cerebral palsy. NeuroRehabilitation 2004; 19: 6978. 9. Palisano RJ, Rosenbaum PL, Walter SD, Russell DJ, Wood EP, Galuppi BE. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Phys Ther 1997; 39: 21423. 10. Russell DJ, Rosenbaum PL, Avery L, Lane M. Gross Motor Function Measure (GMFM-66 and GMFM-88) users manual. Clinics in Developmental Medicine, No. 159. London: Mac Keith Press, 2002. 11. Wood EP, Rosenbaum PL. The Gross Motor Function Classication System for Cerebral Palsy: a study of reliability and stability over time. Dev Med Child Neurol 2000; 42: 29296. 12. Morris C, Galuppi BE, Rosenbaum PL. Reliability of family report for the Gross Motor Function Classication System. Dev Med Child Neurol 2004; 46: 45560. 13. Russell DJ, Rosenbaum PR, Cadman D, Gowland C, Hardy S, Jarvis S. The gross motor function measure: a means to evaluate

the effects of physical therapy. Dev Med Child Neurol. 1989; 31: 34152. 14. Russell DJ, Avery L, Rosenbaum PR, Raina P, Walter SD, Palisano RJ. Improved scaling of the gross motor function measure for children with cerebral palsy. Phys Ther 2000; 80: 87385. 15. Avery L, Russell D, Raina P, Rosenbaum P, Walter S. Rasch analysis of the Gross Motor Function Measure: validating the assumptions of the Rasch model to create an interval measure. Arch Phys Med Rehabil 2003; 84: 697705. 16. Pinheiro J, Bates DM. Mixed-effects models for S and S-Plus. New York: Springer, 2000. 17. Davidian M, Giltinan DM. Nonlinear models for repeated measurement data. London: Chapman and Hall, 1995. 18. S-Plus. Enterprise Developer Version 7.0.6 for Microsoft Windows. Seattle, WA: Insightful Corporation, 2005. 19. Burnham KP, Anderson DR. Model selection and multi-model inference: a practical information-theoretic approach. 2nd edn. New York: Springer, 2002. 20. Williams J. Children with disability grow into adults with a disability. Dev Med Child Neurol. 2008; 50: 163. 21. Bottos M, Gericke C. Ambulatory capacity in cerebral palsy: prognostic criteria and consequences for intervention. Dev Med Child Neurol. 2003; 45: 78690. 22. Pimm P. Cerebral palsy: a non progressive disorder? Educat Child Psychol 1992; 9: 2733. 23. Jahnsen R, Villien L, Stanghelle JK, Holm I. Fatigue in adults with cerebral palsy in Norway compared with the general population. Dev Med Child Neurol 2003; 45: 296303. 24. Murphy KP, Molnar GE, Lankasky K. Medical and functional status of adults with cerebral palsy. Dev Med Child Neurol 1995; 37: 107584.

8 Developmental Medicine & Child Neurology 2009