No Increased Risk for Scoliosis After Early Selective Dorsal Rhizotomy - A Longitudinal Population-Based Controlled Registry Study from four to 25 years of Age

Spasticity interfering with gross motor development in cerebral palsy (CP) can be reduced with selective dorsal rhizotomy (SDR). Although reported, it is unknown if SDR surgery increases the risk for later spine problems. Using CP-registry data from a geographically dened population with the same health care and habilitation services, the objectives were to compare reported scoliosis and spinal pain up to adult age in all SDR-operated with all non-SDR-operated individuals with same medical history, functional abilities, and level of spasticity at four years of age. Method In the total population with CP spastic diplegia in Skåne and Blekinge, born 1990-2006, 149 individuals had moderate to severe spasticity and no medical contraindications against SDR at four years of age and were included; 36 persons had undergone SDR at a median age of 4.0 years (range 2.5 – 6.6 years), and 113 individuals constituting the control group, had not. Data on scoliosis and spinal pain at 10, 15, 20 and 25 years of age were analyzed using Kaplan-Meier survival curves and Fisher’s exact test. Gross motor function classication (GMFCS) levels at four years of age (or pre-operatively) were used for stratication. Result Presence of scoliosis at 15, 20, and 25 years of age was the same in the SDR group as in the control group (p=0.734, 0.735 and 1.0). In severe functional disability (GMFCS IV), the SDR group had later onset and lower occurrence of scoliosis (p=0.004) than the control group. Frequency of reported spinal pain did not differ between the groups. Conclusion nor SDR by natural history. a rest found none


Introduction
Selective dorsal rhizotomy (SDR) is a neurosurgical procedure for children with spastic diplegic cerebral palsy (CP) that permanently reduces spasticity in the lower limbs by cutting parts of lumbosacral rootlets at spinal levels L2-S2. SDR is always combined with physical therapy, and it is mainly used in young children to improve future functional skills (1). Neurosurgeons use either multilevel or single level laminotomy surgery to access the rootlets. As the intervention includes surgery to the spine and spinal nerve roots, there is a hypothetical risk that SDR will cause spinal deformities and pain may develop.
After SDR, scoliosis was reported to be the most common spinal deformity occurring at a weighted mean incidence of 31.6% (2), and Cobb angles ≥ 20° have been reported at 3-9% 2.8-11.6 years after SDR (3)(4)(5). In another study of children with pre-existing scoliosis prior to SDR, 5% improved Cobb angle, 70% were unchanged, and 25% had worsened at a mean follow-up time of 4.3 years (6). Peter et al. and Langerak et al. followed the same ambulatory cohort over time, and at the ve-year follow-up after SDR, scoliosis of ≥ 10° was found in 16% at 4.5 years and 57% at 21 years follow-up, of which 7% had Cobb angles of ≥ 30°. None of the participants had scoliosis prior to SDR (7,8).
On the other hand, persons with CP are generally more frequently affected by spinal abnormalities, particularly scoliosis, than the general population (13).
Follow-up studies of adult individuals walking with or without devices, who had undergone SDR during the preschool years to early school age, show occurrence of spinal pain in 17-23% (7) and pain in spine and lower limbs in 28% (12). In a follow-up with shorter time frame, spinal pain was reported at a lower level, 5% in an self-ambulant cohort 5.8 years post SDR (11).
In studies excluding persons treated with SDR, incidence of spinal pain in CP was shown to increase with increasing scoliosis severity, and scoliosis incidence consequently increased with severity of gross motor dysfunction (14,15).
Today it is unknown whether the above gures regarding spine problems after SDR actually differ from the situation in persons with the corresponding type and severity of CP who are not treated with SDR. In a recently published review of spinal deformities after SDR, the authors suggested that the occurrence of spinal deformities were most likely not different, or only minimally higher after SDR, than in the natural history of CP, but the evidence base was weak (2). They concluded that incidence of spinal deformity after SDR is an important complication to consider, when deciding whether to proceed with the procedure, as spinal deformity can have a major impact on quality of life (3).
The objectives of the present study were to compare presence of scoliosis and spinal pain in individuals who had undergone SDR, and in a control group with same CP-type, medical history, body structure and functions at four years of age, living in the same geographically de ned total population, followed from birth to adult age in the same structured program and registry.

Methods
A retrospective population based controlled long-term outcome study using data from the Swedish cerebral palsy follow-up program (CPUP) including prospectively registered results from repeated, most often yearly, clinical assessments in children, adolescents and adults with CP, living in a de ned geographical area in southern Sweden (16).
The prevalence of CP in the study area was 2.7/1000 children 4-11 years of age as of Jan 1, 2002, of which 38% was spastic diplegia (18). The present study was based on all persons born 1990-2006, with CP spastic diplegia, who lived in Skåne-Blekinge for at least two years during the years 1994-2015; there were 267 persons participating in the CPUP-program and seven (2.5%) who did not. Demographic, medical and functional characteristics of the population with CP spastic diplegia in the study area, and the distribution of these characteristics in the study groups are presented in Table 1.

SDR group
Of the 267 CPUP participants with spastic diplegia, there were 36 persons who had undergone SDR surgery at a median age of 4.0 years (range 2.5-6.6 years). They were followed to a median age of 22.3 years (range 11.3-27.2); 30 of the 36 SDR-operated persons had CPUP assessments at 20 years of age; eight had reached and been assessed at 25 years of age at the end of the study, January 1st, 2018 (Table 1).

Excluded
Of the remaining 231 persons with CP spastic diplegia in the study population, there were 55 persons with medical contraindications to SDR; congenital malformations or syndromes in 25 persons, severe perinatal asphyxia in 17, post-neonatal brain injuries in eight, and ve persons had other severe somatic disorders.
Classi cation of spasticity level at four years of age, as described below, was mild in 63 persons, without any indication for SDR, and they were therefore excluded from the comparison (Table 1).

Control group
The remaining 113 persons with CP spastic diplegia, moderate-severe spasticity level and no medical contraindications against SDR constituted the control group ( Table 1). As of January 1, 2018, this natural history control group was followed to a median of 19.6 years of age (range 8.9-27.3 years).
Six children treated with ITB were excluded from the control group after the ITB-operation at 5, 6, 11, 14, 16 and 18 years of age respectively. In addition, two persons in the control group showed to be too young (< 8 years of age) at the latest assessment and were not included in the comparison. In the control group, 54/113 had reached and had assessments at 20 years of age, and 16/113 at 25 years.

De nitions and classi cations
Cerebral palsy (CP) was de ned according to Much et al. 1992 (21) and the Surveillance of Cerebral Palsy in Europe (SCPE) (22).
CP subtype. Subtypes were de ned by the dominating neurological symptom between four and seven years of age. All patients in this study had Bilateral Spastic CP (BSCP) as de ned by the SCPE, and all also ful lled the de nition of the subtype spastic diplegia in the Hagberg classi cation (18,22). CP spastic diplegia includes all ve GMFCS-levels, as long as there is greater involvement of the legs than of the arms, and can be asymmetrical, with more involvement of either left or right side of the body (23).
Brain morphology was classi ed based on the dominating pattern on brain imaging according to the SCPE (24).
Severe asphyxia was classi ed based on low Apgar scores (< 4 at ve minutes), clinical and EEG-veri ed seizure activities within the rst 72 hours after birth.
Epilepsy was de ned as having had at least two unprovoked seizures after the neonatal period.
Intellectual disability was assessed at about four years of age. Only parts of the cohort were formally IQtested at that age, and only moderate to severe intellectual disability was de ned as intellectual disability in this paper. For the majority of the children, this is a clinical estimation with inherent uncertainty.
The Gross Motor Function Classi cation System (GMFCS) (25) was used to classify gross motor function at each PT assessment in CPUP. The GMFCS level classi ed at the rst assessment after the child´s fourth birthday, or pre-operatively if earlier SDR surgery, was used. The GMFCS was used for classi cation of gross motor function in the CPUP from the year 1995 by some physiotherapists as part of the development/testing of the classi cation, and from 1998, the classi cation was introduced for all PT assessments in CPUP and have shown good stability in previous studies in this population (26).
The GMFCS-level at four years of age was classi ed by the child's physiotherapist in 239 of all 267 persons with BSCP. In the oldest cohort, the GMFCS level was classi ed using other entered CPUP-data at four years of age; the reliability of such a retrospective classi cation was rst con rmed, see Appendix.
The classi cation was performed on de-identi ed data with no information available regarding if the child belonged to the control or the SDR group.
Muscle tone was assessed according to the Bohannon Modi ed Ashworth Scale (MAS) (27). MAS-scores for ve muscle groups (hip-exors, hip-adductors, knee-extensors, knee-exors, and plantar-exors) in both legs were summed up for each CPUP participant with BSCP, who had complete recordings from all included muscle groups at four years of age. The MAS scores 1 and 1 + were both counted as 1. Median MAS summation score was 7 (range 0-32, inter-quartile range [4][5][6][7][8][9][10][11][12]. Muscle tone assessed with MAS is closely related to GMFCS level (28). When occasional single MASscores were missing in the study groups, imputation was therefore performed by using the median score for the missing speci c muscle group(s) of all with completely recorded MAS scores and the same GMFCS level. Six children in the SDR group and 12 in the control group had imputed single MAS scores.
Estimated spasticity level at four years of age was classi ed based on muscle tone (MAS summation score) and clinical signs, such as degree of leg scissoring (none, mild, severe) in rest and in activity, and foot clonus (yes/no), as assessed by each child's local PT according to the structured follow-up form in CPUP (http://cpup.se/in-english/manuals-and-evaluation-forms).
Spasticity level was de ned as mild if the MAS summation score was within the rst quartile 0-3, moderate if in the second quartile 4-6, and severe if summation score was 13 or higher. In the third quartile, MAS summation scores 7-12 the spasticity level was de ned as severe if combined with severe scissoring in activity, and if not, it was classi ed as a moderate spasticity level.
Scoliosis was assessed by inspection by the physiotherapist and registered to be either present or not. If present, mild: scoliosis visible only when leaning forward with aligned pelvis, moderate: scoliosis visible both in leaning forward in sitting and when sitting straight, and severe: scoliosis, not correctable, with need of support in sitting or standing. The CPUP PT assessment of the spine has shown to have high inter-rater reliability and speci city in screening for moderate to severe scoliosis in this population (29).
In the present study, scoliosis status was dichotomized: no scoliosis if there was no scoliosis at examination or when a scoliosis was assessed as being correctable, and existing scoliosis if scoliosis was assessed by the PT as being severe or moderate and not correctable, or if the individual had been undergoing scoliosis surgery. Dates/age at rst PT assessment with existing scoliosis and at scoliosis surgery were noted.
Questions about experienced pain and localization of pain were introduced in the PT assessments from January 2007 (30). Presence of pain was assessed by either the person him/herself or by proxy. For children (< 18 years) pain in the spine was reported by answering yes/no and for adults (> 18 years) recalling the last four weeks for spinal pain that was classi ed as being moderate or severe. After 2016, persons younger than 18 years were also asked to recall presence of pain during the last 4 weeks.

Statistics:
Kaplan-Meier survival curve analyses were performed to explore any differences between rst notation of scoliosis in the registry in the SDR and control group.

Results
The background factors were similar in the SDR group and control group (Table 1). A higher proportion were SDR-operated in the rst two birth cohorts than in the younger cohorts born from 1998 (p = 0.043). Another difference was a lower proportion of severe spasticity in the control group than in the SDR group (p = 0) ( Table 1). Only two of the SDR-operated children were in GMFCS level I, and neither had scoliosis at the end of the study. Two of the 43 in the control group for GMFCS level I had scoliosis at the end of the study (n.s.). No child in GMFCS V had had SDR-surgery. Two of the seven individuals in the control group GMFCS V assessed at 20 years of age had scoliosis.
In GMFCS levels II-IV, no child had scoliosis before ten years of age. Scoliosis was less frequent and developed later in GMFCS II-III with higher level of motor function, than in GMFCS IV. Kaplan Meier survival curves showed no differences between the SDR group and control group GMFCS II and III, with walking ability at four years of age (p = 0.294 and 0.778 respectively) (Figs. 1-2). In the more severe motor dysfunction level, GMFCS IV, individuals in the SDR group developed scoliosis to a lesser extent and with a later onset than individuals in the control group (p = 0.004) (Fig. 3).
All 12 in the SDR group GMFCS IV were followed to age 15 years, and 11 were followed to age 20 years; one SDR group participant GMFCS IV had scoliosis at age 14 years, and one at age 23 years. Eight of the 23 control group participants GMFCS IV developed scoliosis at 12, 12, 12, 13, 16, 18, 18, and 19 years of age respectively. Six of the 14 in the control group followed to at least 15 years of age had scoliosis before end of study, and all ve followed to at least their 20th birthday had scoliosis before that age (Fig. 3).
Patient's reports of spinal pain at 10, 15, 20 and 25 years of age are presented in Table 2. No statistically signi cant difference was observed between the SDR and the control group (Table 2). Questions regarding pain were introduced in CPUP in January 2007, and therefore pain assessments at 10 and 15 years of age were missing for the oldest cohorts. Younger cohorts had not yet reached the oldest ages at data extraction.
Of all those in GMFCS IV who were followed to the 20 years assessment, one of 12 in the SDR and one of seven in the control group reported spinal pain; both had had scoliosis surgery more than four years earlier, and both had severe spinal pain, interfering with daily activities.

Discussion
This is the rst population based controlled SDR long-term outcome study. By selecting all individuals with a medical background and clinical expression that matched the selection criteria for SDR at baseline (1), an SDR-operated group and a control group from the same geographically de ned total population, of which 98% were included in the registry, were created retrospectively. Population, registry and standard of care issues The CPUP registry data provided information from standardized and regular assessments performed and recorded by the person's own physiotherapist at certain ages, same procedures for all individuals, regardless of whether they had undergone SDR surgery or not (16).
All persons in the study were treated at the same public health care units; orthopedic and pediatric hospital health care departments, and the habilitation services in cooperation. Physiotherapy, occupational therapy, social and psychological support, orthoses, braces, orthopedic surgery, ITB, SDR and from 1998 botulinum toxin injections were part of standard care, and with no or very low economic costs for the patients.
Almost all children with CP spastic diplegia in the area were included. Ten children enrolled in CPUP after SDR; their preoperative GMFCS level and muscle tone was assessed and recorded in the medical records by the spasticity-team physiotherapist. All other assessments were performed and recorded in CPUP by the person's local (re)habilitation personnel, which made it possible to study effects of SDR in this population, without any selection bias or bias regarding expectations on SDR-results.

Natural history control group
When SDR was introduced in Lund 1993, three North American randomized controlled trials (RCTs) were underway and preliminary results were forthcoming (32). At the time, an RCT in the Lund university hospital uptake region was considered both unethical and non-feasible due to the small population eligible. Even if the available RCTs showed promising short-term results, monitoring of long-term effects of SDR was needed. In addition to practice-based follow up (33,34), the CPUP program/registry was planned at SDR-start, and among the aims were to follow the natural history and relation to long-term results of treatments (17). The program was started in Skåne and Blekinge 1994 and included persons with CP born in 1990 and later, followed regularly since that time.
The CPUP registry data indicated that only few of the children who could have bene ted from an SDR were referred to the spasticity clinic during early childhood, especially after treatment with botulinum toxin was introduced in 1998. For this study, we therefore could create a control group with clinical background and physical expression to match the selection criteria for SDR (35). Some may have had features not visible in the registry data that differed from those who actually underwent SDR, such as dependence on spasticity for walking and standing, or other barriers to reach the desired functional goals with the intervention. There were also some children included in the control group who were recommended SDR by the spasticity team, but their parents did not choose the intervention.
The pediatric spasticity team members at Skåne University Hospital have had the same selection criteria and follow-up procedures since the start in 1993 (20). Contraindications to SDR were exclusion criteria in the present study, such as mild spasticity, malformation syndromes, postneonatally acquired CP, CP due to prenatal/congenital infections, severe birth asphyxia, and dyskinetic, ataxic, unilateral spastic, or mixed CP subtypes.
Periventricular leukomalacia or hemorrhages, often in combination with premature birth, are associated with the CP subtype spastic diplegia, often suited for the SDR-intervention (1). Such white matter brain lesions were present in the majority of both the SDR group (17/20, 85%) and the control group (74/87, 85%) who had had brain imaging (Table 1). Less CNS-imaging in the SDR group than among controls was due to the higher proportion SDR-surgery in the oldest age cohorts, before brain imaging was recommended in CP diagnostic work-up.

Spasticity
Even if the MAS have shown weak psychometrical properties (36) it has been used by physiotherapists for assessing muscle tone in the CPUP follow-up since the start in 1994. To create study groups at baseline that correspond to the selection criteria for SDR, an estimated spasticity level classi cation was performed as described. The MAS is an ordinal scale and does not methodologically allow such calculations, however it estimates a clinically signi cant entity used for classi cation and not for evaluation of interventions. It makes clinical sense that a child with a high degree of muscle tone in all muscle groups will get a high summation score in contrast to a child with less tone, who may show an increase in just distal muscle groups. In our study, we added clinical signs of spasticity to the MAS summation score quartiles, such as leg scissoring at rest and activity, to classify muscle tone increase into mild, moderate and severe estimated spasticity levels. We found clear cut-offs between the mild, moderate and severe spasticity level groups using the described classi cation, and they were retrospectively found to t the overall clinical picture; none of the children in the SDR group ended up in the mild spasticity level group at baseline.

GMFCS
The GMFCS levels were classi ed after and as close to the fourth birthday as possible, and they were used to stratify the study population at baseline. At four years of age, the GMFCS level, CP diagnosis, and CP subtype can be decided with high or acceptable accuracy (22,37).
Retrospective, although not psychometrically tested, classi cation of GMFCS levels based on clinical descriptions in medical records was used in the 2002 metanalysis of the three North American SDR RCTs (32). In the present study, structured data from the CPUP registry on functional performance and capability was available for retrospective classi cation of GMFCS levels at baseline date before the GMFCS was introduced. The kappa analysis of this classi cation showed good agreement (κ = 0.732, p < 0.001), as described in Appendix, and the oldest birth year cohorts could be included in the study.
Each GMFCS level carries a meaningful distinction about function, and it may wrongly lead to the collapse of different GMFCS levels in analyses (38). Therefore, participants at GMFCS levels I and V were excluded from part of the analyses, as it showed that only two and no child respectively at these levels had had SDR-surgery. Also, scoliosis development at each GMFCS level was presented separately.

Scoliosis
The multilevel laminoplasty technique used to access the rootlets for the SDR procedure in the present study included reinstatement of laminae and did not increase the occurrence of scoliosis after SDR. For the SDR GMFCS IV-group, scoliosis occurred even to a lesser extent and with later onset than for GMFCS IV-control group (Fig. 3). Development of contractures and asymmetries, especially common in higher GMFCS-levels (15,39,40), may be less severe after SDR combined with physiotherapy, as use of orthoses, sitting and supported standing positions with more symmetric spine may be more easy to obtain after tonus reduction.
In SDR, other forms of spinal misalignments, especially spondylolisthesis have been reported more frequent than in the general population (4,5,7,9,10), even if scoliosis was reported to be the most common deformity following SDR (2). Studies reporting spinal deformities after SDR are not population based and with no or small comparison groups (2). To further explore spinal misalignments after SDR in the present population, results regarding imaging of the spine beyond scoliosis and Cobb angles would be needed. Absence of spinal pain may, however, indicate absence of signi cant spinal problems.

Pain
Pain in the CP population has previously not been properly noticed (30), even if it is one of the most common co-morbidities (41). Beside no increase in scoliosis development after SDR, the other main nding of the present study was that the frequency of spinal pain did not differ between the SDR group and the control group at 10, 15, 20 and 25 years of age (Table 2).

Page 16/24
The low proportion of individuals born 1990-1997, available to serve as natural history control group in adult age, is a study limitation ( Table 1). The control group thus included a higher proportion of younger persons, who probably received somewhat different care compared to the older cohorts (16,17). In study participants born 1994 and later, in contrast to those born 1990-1993, some were treated with botulinum toxin with a lowered muscle tone at the baseline assessment.
Assessments and registrations were performed regularly using a standardized methodology by clinicians in their daily practice, and limited information was available, as only the most important items can be included to keep a register acceptably time-consuming. The spinal screening lacked information on other misalignments than scoliosis, and x-ray examinations with Cobb angles were inconsistently registered at the time data was extracted from the register.
The late introduction of pain screening in the registry resulted in low numbers of recorded answers about spinal pain at different ages, which is another limitation. There were slightly more frequent recordings of spinal pain in the SDR than in the control group, although the differences were not statistically signi cant. Although possible reduction of pain in adults with CP after early SDR is reported (42), the authors are anxious to nd out whether SDR at a young age is causing more spinal pain than expected from natural history. Information of pain intensity, duration, effect on daily living or quality of life was available only in the adult age CPUP forms.
Spinal pain several years after surgical correction of scoliosis, as described in the present study, was found also in a population based study with high number of participants (14). Increased awareness among health professionals of the importance of pain assessments in this population led to extended pain questions in the most recent version of the CPUP PT-form, so more and higher quality data will be available in the future.

Generalizability
This study represents the real-life situation in the ordinary health care, in contrast to RCTs, usually conducted at tertial health care level after rigorous selection of participants. The total population with CP in certain age cohorts were included in the present study, without selection bias. Results would be generalizable in populations where the socio-economic and health care standards are comparable to those in Sweden. Also, the surgery technique, multi-level SDR without permanent removal of the spinal laminae/spinous processes, was used for all individuals in the study, and is commonly used internationally.

Conclusion
This population based longitudinal matched outcome study, provides evidence against long-term complications from the spine caused by the SDR surgery. Individuals undergoing SDR had similar development of scoliosis as comparable controls. In addition, individuals with most functional limitations, GMFCS IV, who had SDR in young age had later onset and lower occurrence of scoliosis than their peers in the control group. Spinal pain was reported at similar levels for SDR operated and controls up to the age of 25 years. Permission to extract data from the CPUP registry was obtained by the registry holder and the personal data responsible authority (Region Skåne).
The parent or the legal guardian provided oral consent prior to participation in CPUP, and the children provided verbal assent, as applicable. Verbal consent is su cient for participating in the Swedish national quality registries. Participation can be discontinued at any time, and the decision to withdraw will not affect the healthcare received.

Consent for publication
Not applicable Availability of data and materials Kaplan-Meier curve illustrating age when scoliosis (according to the study de nition) was rst reported in GMFCS level III. Legend: Analysis of data from individuals who had undergone Selective Dorsal Rhizotomy (SDR) (n=11) (blue) and the control group (n=20) (green), p=0.778 (n.s.)