Comprehensive short-term outcome assessment of selective dorsal rhizotomy


  • See end of paper for list of abbreviations.

* Correspondence to first author at 200 East University Avenue, St Paul, MN 55101, USA.


This study retrospectively evaluated the safety and efficacy of selective dorsal rhizotomy (SDR) in participants who underwent a rigorous selection process, uniform surgical procedure, and a standardized postoperative rehabilitation process. Outcome measures assessed were the Ashworth scale for spasticity, the Gillette Gait Index (GGI) for overall gait pathology, oxygen cost for gait efficiency, and the Gillette Functional Assessment Questionnaire (functional walking ability scale; [FAQ]) for functional mobility. Outcomes were evaluated for 136 children (81 males, 55 females; mean age 7y 3mo [SD 2y 1mo], range 3y 5mo–18y 9mo) for an average of 18.3 months (SD 4.4mo) postoperatively. All participants had a diagnosis of cerebral palsy (CP): 10 quadriplegia, 19 triplegia, and 107 diplegia. Preoperative Gross Motor Function Classification System levels were: Level I n=6; Level II n=64; Level III n=59, and Level IV n=7. All outcome measures improved for the group as a whole. Spasticity improved with 66 to 92% of possible gain in Ashworth scores; GGI was 7.5 times more likely to have a good as opposed to a poor outcome; energy efficiency improved in over half of the participants, and the FAQ demonstrated a statistically significant improvement of 0.9 levels (p<0.001). The rate of complications was low, with peri- and postoperative complications resolved by time of discharge.

Despite 20 years of use in cerebral palsy (CP), the selective dorsal rhizotomy (SDR) procedure remains controversial.1 This is in part due to the risk of peri- and postoperative complications, and also to concerns about the longer-term possibility of crouch gait, spinal deformity, foot deformity, hip subluxation, and functional weakness.2–4 Even among those who use SDR, patient selection criteria and surgical technique are not uniform.2,5 SDR is generally used in a subgroup of individuals with CP to reduce spasticity and increase function. The selection criteria for that subgroup vary from place to place and even within an institution.2,3,5 Published guidelines exist for age, diagnosis, tone, ambulatory ability, birth history, motor control, specific medical conditions, orthopaedic status, availability of postoperative therapy, and intellectual development.2 However, despite the controversy, SDR has been shown to reduce spasticity, improve range of motion, function, and self-care abilities, and normalize gait patterns.3

The purpose of this retrospective study was to evaluate short-term outcomes and complications of SDR for patients who underwent a rigorous selection process, uniform surgical procedure, and standardized postoperative rehabilitation. This study did not address long-term outcomes or complications of SDR, nor did it compare and contrast SDR to other methods of treatment.


Institutional review board approval was obtained for this retrospective analysis. A waiver was obtained to access protected health information without informed consent. Included in the study were participants who met the following criteria: (1) underwent SDR at Gillette Children’s Specialty Healthcare or Shriner’s Hospital for Children – Twin Cities Unit, Minneapolis, MN between the 30th January 1994 and 1st January 2003; (2) had preoperative gait analysis up to 18 months before SDR and postoperative gait analysis between 8 and 36 months after SDR; (3) had no intervening surgery between gait analysis and SDR other than hardware removal.

Birth history, imaging data, medical history, spasticity, selective control, strength, surgical history, SDR technical details, intra-, peri-, and postoperative complications, gait kinematics and kinetics, gait electromyographic (EMG) data, and oxygen cost were extracted from operative reports, in-patient notes, discharge summaries, and gait analysis reports. An extensive multidisciplinary report that included input from orthopaedics, neurosurgery, pediatric rehabilitation medicine, physical therapy, nursing, and social work was obtained from the Spasticity Evaluation Clinic at the Gillette Children’s Specialty Healthcare.

To ascertain the criteria used to select patients for SDR, interviews were conducted with the clinicians at the Spasticity Evaluation Clinic. These produced the following consensus description of the ideal SDR candidate.

Birth history

Preterm, defined as birth at less than 36 weeks’ gestation. Neuroimaging evidence of periventricular leukomalacia, which is indicative of a brain insult that typically results in spasticity rather than mixed tone abnormalities.


Purely spastic tone, as noted on clinical examination and suggested by computerized gait analysis (repeatable EMG and gait pattern between walking trails and limited sagittal plane range of motion); minimal evidence of mixed tone on a motion assessment physical examination by a physical therapist and physicians; ashworth scores of between 2 and 4 at rectus, hamstrings, hip flexors, adductors, and plantarflexors.6

Strength and selective motor control

Complete or partial ability to isolate movement at the hips, knees, and ankles as measured on physical examination; absence of reflexive movement patterns during walking (i.e. exaggerated hip flexion, knee flexion, and ankle dorsiflexion during swing); strength of at least antigravity for the hip flexors.

Gait inefficiency

Oxygen cost (gait inefficiency) greater than two times that of speed matched controls.

Outcome measures were selected at the impairment and activity levels of the International Classification of Functioning, Disability and Health framework.7 These were Ashworth scale for spasticity6 (impairment), Gillette Gait Index8 (GGI) for overall gait pathology (impairment), oxygen cost for gait efficiency9 (activity), and the Gillette Functional Assessment Questionnaire10 (functional walking ability scale; [FAQ]) for overall ambulatory function (activity).

The Ashworth scale measures the impairment of spasticity that is primarily addressed by SDR. It was assessed by a physical therapist at the time of the pre- and postoperative gait analyses. The 5-point ordinal scale for grading the resistance is: 1=no increase in tone; 2=slight increase in tone, giving a catch when the limb is moved in flexion or extension; 3=more marked increase in tone but limb easily flexed; 4=considerable increase in tone, passive movement difficult; and 5=limb rigid in flexion or extension.6

The GGI is a multivariate measure quantifying the degree of gait pathology and measures how closely an individual’s gait pattern approaches typical gait. The GGI for a typical child is 15, with a 95th centile range of 8.2 to 26.9.8 The GGI has been shown to be valid across sites, and to be clinically relevant.11

Net non-dimensional oxygen cost, which reflects overall gait efficiency, along with dimensionless speed (defined as speed/[gravity*leg length]0.5) were collected at the time of the complete gait analysis on participants able to complete a 6-minute walk. Data are reported as the ratio of participant cost to able-bodied-control cost at matched speed (percent control cost). The net non-dimensional scheme controls for the effects of body mass, height, age, and sex.9

The FAQ is a 10-level parent-reported functional measure of patient ambulation in the patient’s own environment.10 A child who is typically able to keep up with peers achieves a score of 10. The scale decreases with decreasing ability for community ambulation.

Statistical Analyses

A uniform scheme for determining categorical outcomes was used consistently across all outcome measures. The scheme allowed an outcome to be rated as poor (worsening or loss from within typical range), neutral (no change or maintenance within typical limits), or good (improvement or correction to within typical limits).

Pre–post changes were assessed using paired t-tests with a 5% threshold for significance. Patients for whom pre- or postoperative data were missing were excluded from the analysis of the relevant measure. For example, a patient missing preoperative FAQ data would be excluded from the FAQ analysis, but included in assessment of other outcome measures (GGI, oxygen, etc.). Data deemed to be normally distributed (Kolomogorov-Smirnov test) were analyzed using paired t-tests. Otherwise, the non-parametric Wilcoxon signed ranks test was used.

Operative and postoperative procedures

Preoperative sedation was not administered as this alters intraoperative neurophysiology.12,13 Induction and maintenance of general anesthesia was accomplished with desflurane at 4 to 6%, or sevoflurane at 1 to 1.5% and a short acting paralytic agent for endotracheal intubation (no narcotics, benzodiazepines, or succinylcholine). The optimal nitrous oxide concentration is 50% as higher concentrations negatively affect stimulation thresholds. Propofol at normal concentrations was used, as it does not interfere with H-reflex activity.14

With the patient in the prone position, insulated needle electrodes were placed in the bilateral adductors, quadriceps, medial and lateral hamstrings, tibialis anterior, gastrocnemius, and gluteus medius, as well as the anal sphincter, to monitor for response to stimulation.

The dura was exposed by a trapdoor laminaplasty as previously described.12 The dural sac was opened midline and tack-up sutures placed to expose the cauda eqina. Rhizotomy hooks were placed 1cm apart and motor stimulation was accomplished using 0.1mA at 50Hz for 50ms.12,15,16 Once motor level was established, the root was divided into motor and sensory divisions. The sensory roots were micro-dissected (150–250 rootlets), and each component rootlet was stimulated at threshold, beginning at S1, to establish threshold response.12 The S1 starting level was chosen as it typically reflects the highest level of spasticity, and elimination of these rootlets with abnormal response has an impact on electrophysiological recording at subsequent levels.12 The sensory stimulation for threshold response ranged from 0.2mA to 3mA. Care was taken to avoid delivering supra-threshold stimulation.

EMG response and clinical observations were recorded for each rootlet. Single twitch, decremental, or squared responses were considered to be normal. Incremental, clonic, multiphasic, sustained, and responses that spread to three or more muscles beyond the primary level of stimulation or the opposite leg were considered abnormal. Rootlets with abnormal response were cut. This paradigm was repeated for each sensory root on both sides S1 (or S2) to L1. Rootlets that only produced anal sphincter activity were preserved.

Typically, the total number of rootlets cut was between 25 and 45%. The upper limit was never exceeded for the patient, though at any individual level more than 45% of rootlets may have been cut. When the overall number of abnormal rootlets exceeded 45%, clinical examination and gait analysis data were used to decide which of the abnormal rootlets should be cut.

Intravenous narcotics and 0.25% bupivacaine injected into the paraspinal muscles were used intraoperatively for postoperative pain relief. Knee immobilizers were used to reduce postoperative muscle spasms. Ketorolac, narcotics, and muscle relaxants were used for the first 48 hours postoperatively.

Although most studies of SDR indicate that surgery is followed by intensive rehabilitation, what constitutes intensive rehabilitation is often not detailed. Some studies report discharge shortly after surgery, with varying amounts of outpatient physical and occupational therapy, while others indicate an in-patient stay of 4 to 6 weeks.5 Rehabilitation protocols were initiated 3 days after surgery. The children remained in hospital for an average of 40 days, participating in an intensive in-patient rehabilitation program including twice- daily physical and occupational therapy. The program emphasized the mastery of lower level motor skills before progressing to higher level skills, strengthening, and selective control.17,18


Adherence to selection criteria was as follows: explicit adherence (missing data counted as non-adherence): birth history 77%, imaging information 26%, selective control 81%, tone 76%, strength 87%, and energy cost 60%. Neutral adherence (missing data ignored): birth history 88%, imaging information 30%, selective control 98%, tone 78%, strength 99%, and energy cost 85%.

A mean of 36% of rootlets were cut in this group of 81 male and 55 female participants (Table Ia). The majority of participants in the study were in GMFCS Levels II (n=64) and III (n=59), with only six GMFCS Level I and seven GMFCS Level IV participants meeting the clinical eligibility criteria for SDR (Table Ib). The most common diagnosis was diplegic CP (n=107), with fewer quadriplegic (n=10) and triplegic (n=19) participants. Peri- and postoperative complications were infrequent, and resolved by discharge (Table II). At the 1-year follow-up no documented spinal problems existed.

Table Ia.   Participant profile
SexNumberAgeFollow-up time [mo]Percent rootlets cut
Female557y 7mo2y 1mo19.1 (4.9)35
Male816y 11mo2y 0mo17.8 (4.0)37
Total1367y 3mo2y 1mo18.3 (4.4)36
Table Ib.   Functional profile
SexNumberGMFCS – Pre SexGMFCS – Post -SDR
  1. GMFCS, Gross Motor Function Classification System.

Table II.   Transient complications
Bowel and bladder118
Skin related97
Wound healing86
Miscellaneous related54
Miscellaneous not related32

All outcome measures improved for the group overall (Table III, Fig. 1). Every participant underwent gait analysis, and thus had pre- and postoperative GGI scores. Ashworth scores were available for 99% of participants, energy cost for 71%, and FAQ scores for 89%.

Table III.   Outcome measures – group quantitative response
MeasureParticipants n%Pre-SDR mean (SD)Post -SDR mean (SD)TypicalChange mean (SD)% Possible gain pTest
  1. SDR, selective dorsal rhizotomy; GGI, Gillette Gait Index; FACD, Functional Assessment Questionnaire

GGI136100242 (123)180 (128)15−62 (105)27<0.001Paired t
FAQ121897.3 (1.7)8.2 (1.3)100.9 (1.5)33<0.001Wilcoxon
Oxygen Cost (% typical)9671343 (149)291 (142)100−53 (132)22<0.001Paired t
Dimensionless Walking Speed96710.26 (0.11)0.29 (0.09)0.430.03 (0.07)27<0.001Paired t
Ashworth ScoresN sides        
 Hip Flexors133981.6 (0.8)1.1 (0.3)10.5 (0.8)83<0.001Wilcoxon
 Hip Adductors134992.8 (0.8)1.3 (0.6)11.4 (0.7)78<0.001Wilcoxon
 Hamstrings134992.1 (0.7)1.1 (0.4)11.0 (0.8)91<0.001Wilcoxon
 Rectus Femoris135992.5 (1.1)1.3 (0.6)11.2 (1.1)80<0.001Wicoxon
 Plantarflexors135993.0 (0.8)1.6 (0.6)11.3 (0.9)65<0.001Wilcoxon
Figure 1.

Categorical outcomes for the spectrum of measures included in this study are shown. These include measures of overall function (FAQ), spasticity (Ashworth scores), gait efficiency (oxygen cost), and gait pattern pathology (GGI). Good outcomes range from around 30% (FAQ) to 70% (GGI). There is a very low rate of Poor outcomes (≈ 10%). FAQ, Gillette Functional Assessment Questionnaire (functional walking ability questionnaire); GGI, Gillette Gait Index.

Changes in GGI were largely due to improvements in the sagittal plane [Fig. S1, Supporting Information published online). Mixed results were seen at the pelvis, with an improvement in the ‘double-bump’ pattern but a worsening in mean tilt. Hip range-of-motion and maximum extension improved, as did the knee flexion/extension curve (initial contact, loading response, maximum extension, range-of-motion, peak swing). At the ankle there was significant reduction of equinus and improved consistency of kinematic patterns. Overall, 71% of patients exhibited a good outcome, 16% neutral, and 12% poor outcome.

Spasticity decreased substantially at the major muscle groups tested (Table SI, Supporting Information published online). A large percentage of muscle groups with preoperative spasticity had complete correction (postoperative Ashworth score 1): hip flexor (83%), adductor (72%), hamstring (84%), and rectus femoris (74%). Complete correction of plantarflexor spasticity was not as common (41%).

While the oxygen cost decreased for the group as a whole, there was heterogeneity of the response [Fig. S2, Supporting Information published online]. The predominant response was one of increased speed and decreased oxygen cost.

Overall ambulatory function (FAQ) improved for many participants in the study. Few participants exhibited a decrease in FAQ (Table SII, Supporting Information published online). Participants with a preoperative FAQ range of 6 to 7 demonstrated a mix of neutral and good responses. Participants who had a preoperative FAQ of 8 or higher had predominantly neutral responses owing to a ceiling effect, and the fact that a 1-level improvement was categorized as ‘unchanged’. The only way for these participants to achieve a ‘good’ outcome was to reach FAQ level 10.

Categorical outcomes were stratified by preoperative GMFCS level (Table IV). Because the number of participants in GMFCS Levels I and IV was small, further statistical analyses were conducted on participants in GMFCS Levels II and III. χ2 analyses of good-neutral-poor outcomes showed that only speed outcomes differed between the groups (p<0.01), with more good outcomes among GMFCS Level III participants and more neutral outcomes among GMFCS Level II participants. It should be noted that GMFCS Level II participants walked significantly faster preoperatively (inline image), limiting possible gains in speed among these higher functioning children.

Table IV.   Outcome
MeasurePre-op GroupGood OutcomeNeutral OutcomePoor Outcome
  1. Numbers of study participants with outcomes in respective categories.

SpeedGMFCS I   4 1  
GMFCS II 7 33 111
GMFCS III1012 9 932
GMFCS IV 1 1  1  
O2 costGMFCS I  4  1  
GMFCS II 623 1031
GMFCS III 120 1383
GMFCS IV  1  1 1
GMFCS II 444 124 
GMFCS III 43 106 
GMFCS IV  3  4 
FAQGMFCS I 1  3 2  
GMFCS II10 9 351 
GMFCS III 216 314 
GMFCS IV  5  2  


SDR is an irreversible procedure; therefore careful analysis of outcomes to assess safety and efficacy is necessary. This study described explicit patient selection criteria and adherence to these criteria, described the surgical technique in detail, completely documented intra-, peri-, and short-term postoperative complications, and measured patient outcome in a comprehensive manner.

Adherence to the selection criteria was good, with the exception of imaging data which was poor, and energy cost which was average. Publication of these criteria is intended to describe the methodology of this study and assist others in evaluating selection criteria. Due to a number of factors, it is not clear whether failure to adhere to these criteria poses a risk to patients (e.g. small number of poor outcomes, small number of SDRs performed on children who did not meet the criteria, missing data). Previously published selection criteria tend to be general with no analysis of adherence.19–21 van Schie et al. describes the most specific selection criteria for a study of nine children.22

Using the surgical technique described here, the rate of intra-, peri-, and postoperative complications was low (Table II). All complications listed resolved by the time of discharge (6wks postoperatively), indicating a high level of technical safety of the procedure.23 Careful intraoperative monitoring, meticulous hemostasis, and surgical decisionmaking limiting the percentage of rootlets cut, seemed to contribute to the safety and efficacy of the procedure.

Long-term complications were not analyzed in this study. Several other studies stress the importance for orthopaedic follow-up after SDR to treat spinal deformities, hip subluxation, lever arm dysfunction, foot deformities, and residual muscle tightness.24 Spiegel et al. documented long-term spinal deformity outcomes (mean follow-up 4y 2mo, SD 2y) in a comparable set of participants and found a 17% incidence of scoliosis and a 13% incidence of spondylolisthesis.25 As noted by Spiegel et al., the high rate of scoliosis in CP and the lack of historical controls compromise the ability to interpret the findings. Carroll et al. reported that 37% of their participants required operative intervention for planovalgus foot deformities.26 The percentage of rootlets cut averaged 50%, with 98% having hypotonia after SDR. As is the case with scoliosis, foot deformities are widely observed in CP. The short-term nature of this study rendered an analysis of subsequent orthopaedic surgery impossible and inappropriate. However, experience has shown that many of these patients will require orthopaedic intervention at a later date.27

The multiple domains of outcome show a high level of efficacy for SDR performed in the described manner, on participants meeting the described selection criteria. SDR was shown to be effective in reducing spasticity, improving gait quality, reducing energy cost, and improving overall ambulatory function.

Spasticity improved substantially for each muscle group, with 65 to 91% of possible gain in Ashworth scores (Table III, Fig. 1). While it may be possible to reduce spasticity further by sectioning a higher percentage of rootlets, it appears that low levels of residual spasticity in children with mildly impaired motor control may be functionally beneficial for joint stabilization and power production. Relief of spasticity is not an end in itself, but rather a means to improve function.

Gait pathology, as defined by the GGI, was about 7.5 times more likely to have a good rather than a poor outcome (71%:13% good:poor outcomes; Table IV). These post-SDR improvements are believed to be attributable to the reduction in spasticity. Tone reduction achieved via SDR affects all joints and planes to some degree. However, the most notable changes occurred in the sagittal plane, and for this reason these changes are featured (Fig. S1). Some improvements were in range of motion and/or modulation, while others were manifest as normalization in timing of key gait events. It should be noted that even post-SDR, the GGI remained significantly elevated compared with unimpaired gait. Orthopaedic deformity, weakness, and deficits in motor control contribute to the residual gait pathology.

An elevated oxygen cost is common in children with CP. Improved energy efficiency was seen in over half of the participants (Table IV; Figs. 1 and 4). Spasticity, weakness, orthopaedic deformity, and impaired motor control also contribute to energy inefficiency. These factors probably account for the 25% of patients who showed no change in oxygen cost, as well as the residual elevated energy cost in many participants.

In general, SDR improved functional ambulation. The use of the 10-level FAQ demonstrated a statistically significant improvement of 0.9 levels (p<0.001; Table III). Those who were limited community ambulators preoperatively (FAQ level 6 or 7) often improved postoperatively to become good functional community ambulators (FAQ level 8 or 9). The majority of patients who were good functional community ambulators preoperatively (FAQ level ≥ 8) did not change by more than one level (ceiling effect; Table SII).

Patients are not chosen for SDR based on their GMFCS level. However, the relationship between GMFCS level and strength, spasticity, and oxygen cost implicitly results in a predominance of children in GMFCS levels II and III receiving SDR at the Gillette Children’s Specialty Healthcare (Table IV). Children who used assistive devices preoperatively (GMFCS Levels III/IV) had an equal or better likelihood of a good outcome for gait (GGI, speed respectively) and ambulatory function (FAQ). Of the children who used assistive devices preoperatively, 32% (21/66) no longer used devices postoperatively. Conversely, only 3% (2/70) moved from independent to device-dependent walking. Children who used assistive devices for community ambulation were more likely to show improvements in speed postoperatively than their community ambulating counterparts, due, in part, to the fact that the independent ambulators walked faster preoperatively.

There was no control group for this study. Children meeting the same selection criteria, who did not undergo SDR were not analyzed. In a broader group of children Schwartz et al. found that children who do not undergo SDR before orthopaedic surgery have substantially higher rates of soft tissue surgery than those who have had SDR.27 This reinforces the notion that spasticity reduction and orthopaedic surgery are complimentary elements of an integrated management approach.

Intrathecal baclofen (ITB) pumps are known to be effective spasticity reducers. While ITB users report high levels of satisfaction, and exhibit significantly reduced hypertonia, there is little evidence that an ITB improves gait.28 ITB treatment does improve some dimensions of the GMFM for children in GMFCS Levels III, IV, and V.28 At Gillette Children’s Specialty Healthcare, ambulatory children considered for ITB do not generally meet the criteria for SDR. As a result, ITB and SDR patients are not comparable at this institution.

Study Limitations

There was no control group for this study. Children meeting the same selection criteria, who did not undergo an SDR were not analyzed. The short-term design was specifically aimed at evaluating the effects of SDR, eliminating the effects of growth and other interventions. This is both a strength and a weakness. This study did not control for the effects of rehabilitation during postoperative recovery. In-patient rehabilitation is standardized, but variations inevitably exist. There is a uniform recommendation for outpatient rehabilitation, but specific rehabilitation was neither controlled for nor assessed in this study.


SDR is safe and effective when performed in the manner described in this study, on patients that match the stated selection criteria. Relatively few poor outcomes were observed (10% or less for each measure), and no participant had a poor outcome on every measure. There may be participants who could benefit from SDR but who did not receive treatment. Future studies will examine data from patients who were considered for, but did not undergo an SDR. This study demonstrated positive short-term effects of SDR; long-term outcomes remain a primary focus of future SDR research.

List of abbreviations

Gillette Functional Assessment Questionnaire


Gillette Gait Index


Selective dorsal rhizotomy