Gastrocsoleus Lengthening in Cerebral Palsy

The objective of the current study is to analyze the development of ankle and knee range of motion (ROM) related to type and level of surgery, age at surgery and CP-subtype.

Materials and methods

Scope of the analyses The analysis will compare knee and ankle ROM development between types and levels of surgery performed to achieve lengthening of the gastrosoleus complex. Also, associations between treatment outcomes and GMFCS-level as well as CP-subtype will be evaluated.

Study Objectives To determine which, if any, surgery at muscular level (zone 1 and 2), open surgery at tendinous level (zone 1) and percutaneous surgery at tendinous level (zone1) is more advantageous in terms of subsequent knee and ankle ROM development. We hypothesize that children with BSCP having surgery on the Achilles tendon is more prone to develop exaggerated dorsiflexion in the ankle as well as knee contractures.

Endpoints

Primary endpoints

• Polynomial regression coefficients describing mean ankle ROM development over follow-up time.

Secondary endpoints

Polynomial regression coefficients describing mean knee ROM development over follow-up time.

Composite endpoint: the proportion of patients with failure is defined as dorsiflexion of the ankle ≥ 20°, or dorsiflexion of the ankle ≤ 0 degrees defining recurrent equinus. Subjects will be followed from date of surgery to failure, the last assessment or end of study, 31 dec 2022.

Each component endpoint: Dorsiflexion of the ankle ≥ 20° dorsiflexion of the ankle ≤ 0 degrees or flexion contracture in the knee ≥10°, . Subjects will be followed from date of primary surgery to failure, the last assessment or end of study, at year 2022.

General Study Design and Plan

This is a retrospective longitudinal superiority study on the effects of gastrocsoleus complex lengthening on ROM development. The study is based on data from the Swedish Surveillance Program for Cerebral Palsy (CPUP) registry that covers >95% of children with CP in Sweden(Alriksson-Schmidt et al. 2017). In CPUP the diagnosis is confirmed at the age of 4 by a neuropaediatrician (Westbom et al. 2007, Hollung et al. 2020). All families and guardians had given informed verbal consent to use data for research before inclusion in the CPUP registry. In Sweden the multidisciplinary habilitation units are responsible for children with CP and the child's physiotherapist perform regular assessments of clinical variables necessary to evaluate the health status of the child, including range of motion (ROM). Assessments are performed twice a year, once a year or every second year, depending on the GMFCS-level and age. The assessments are performed in standardized position of the child and goniometer arms according to the CPUP-manual (http://www.cpup.se).

Children treated with gastrocsoleus lengthening and at least 7-year follow-up will be collected from the CPUP registry. Type and level of surgery performed will be validated from the patient records of the child. Treated children will be stratified into 3 groups depending on the type and level of gastrocsoleus surgery: (surgery at muscular level (Zone 1 and 2), versus open surgery at tendinous level (Zone 1), versus percutaneous surgery at tendinous level (Zone1)). Groups are then followed for a minimum of 7 years (last date of surgery December31, 2014) or until possible reoperation with respect to ROM development. Depending on the distribution of GMFCS levels of each child, this will entail a varying number of measurements but typically around 8-10.

Between-group comparisons will be performed using mean ROM measurements over the entire follow up by means of statistical modeling and hazard rate ratios (HR) and 95% confidence intervals for complication rates.

Inclusion-Exclusion Criteria and General Study Population

All children born 2000 - 2011 who accepted inclusion into the CPUP follow-up program and registry at any age from the year 2000 and onwards, with an isolated gastrocsoleus complex lengthening performed between 2000 and 2013, to allow for at least 7 years of follow-up.

Children treated with other surgical procedures (soft tissue surgery to the foot, leg, knee and hamstrings and osteotomies or arthrodesis to all lower extremity) or where the type of gastrocsoleus lengthening could not be verified, will be excluded from the analysis but described using summary statistics.

Children with no ROM measurements before or after surgery will be excluded as well as children treated with intrathecal baclofen and dorsal selective rhizotomy.

Study Assessments

Frequency of ROM - measurements in CPUP: Children in GMFCS I < age 6 is assessed once every year and children < age 6 in GMFCS II-V twice every year. Children in GMFCS I > age 6 is assessed once every second year and children > age 6 in GMFCS II-V once every year.

Variables

Type and level of surgery (TLS) - Collected from the CPUP registry and validated through patient records. Takes on the values "TAL", "Open TAL", "Gastrocsoleus lengthening in zone 1 and 2".

Two indicator variables derived from the TLS variable; one indicator for the "Open Z" group, and one for the "Zone 1 and 2"; they take on the value 1 when a child is part of the group and 0 when it is not.

GMFCS level - ordinal scale of levels I through V. Type of spastic CP - indicator variable taking on the value 0 for USCP and 1 for BSCP and is empty for any other CP type.

Age at surgery - derived as the difference in years between date of surgery and birth date.

ROM measured by means of goniometer

Knee extension, takes on values between 0 and -90 Ankle dorsiflexion, takes on values between -45 and 45 Three dichotomized failure variables

An indicator variable for knee extension ≤ -10° An indicator variable for ankle dorsiflexion ≥ 20° An indicator variable for ankle dorsiflexion ≤ 0 Spasticity measured according to the Ashworth scale 0 to 4(Bohannon et al. 1987).

Sample Size

As this is a register-based study the sample size is set by design. From reviewing register data, we guestimate that about 200 children will fulfill the inclusion criteria. With an approximate standard deviation in ankle dorsiflexion of 11 degrees we will be able to detect a mean difference between groups of 3.1 degrees, at the alpha level of 0.05 for a two-sided T-test, with 80 percent power. This is below the minimal clinical important difference (MICD) as well as below the measurement error of the goniometer, i.e. 5 degrees(Konor et al. 2012). For failure rates, we guestimate that there will be approximately 100 events for our secondary composite endpoint, which means we will be able to detect a hazard rate ratio of 1.8 at the alpha level of 0.05 for a log rank test, with a power of 0.8.

Timing of Analyses

Final analysis of the data will be initiated when

Data has been extracted from the CPUP register, including all visit up until the day of extraction.

Type and level of surgery has been validated by a subject matter expert using patient records.

Remaining data has been validated and cleaned by a subject matter expert. Data has been transferred to an analysis file and submitted to the statistician.

Covariates and Subgroups

Covariates Covariates assumed to influence the primary and secondary endpoints are sex, age-at-surgery, baseline knee and ankle ROM, baseline spasticity level: Ashworth, GMFCS level and CP-subtype: unilateral spastic cerebral palsy (USCP) vs bilateral spastic cerebral palsy (BSCP).

Confounders The set of covariates used for confounding adjustment is selected according to the disjunctive cause criterion. Here, covariates that influence outcome and/or exposure, i.e. choice of surgery, (age-at-surgery, baseline ROM, baseline spasticity, GMFCS-level) and covariates assumed to influence exposure, i.e. choice of surgery, (age-at-surgery, baseline ROM, baseline spasticity, GMFCS-level and subdiagnosis) will be included in the final model.

Subgroups Surgery results are hypothesized to depend on the specific CP-subtype of the child, operationalized in the current study as either USCP or BSCP. Therefore, interaction analysis intended to identify modifiers of treatment effects are performed using the CP-suptype, grouped accordingly . However, this analysis will be exploratory in nature, as the study is likely underpowered to detect these interaction effects.

Missing covariate data will be ignored if, when combined over all covariates, the proportion missing in any covariate constitute less than 5 percent. If the missing proportion exceeds 5 percent, a sensitivity analysis will be performed using multiple imputation by chained equations imputing missing covariate and outcome data.

Multiple Testing

No adjustment for multiplicity will be performed in the current study, as it is observational and exploratory in nature(Bender et al. 2001).

Derived Variables

Two variables will be derived from study data

The primary surgical grouping variable will be created based on data extracted from the CPUP register, where surgical information is recorded including date and type of surgery. Here all data on all visits up until the day of extraction will be included. Type and level of surgery have recently been validated. The patients will be divided into 3 subgroups depending on the type and level of gastrocsoleus surgery.

The outcome variable, "failure", used for the secondary composite endpoint; "the proportion of patients with failure", will be created using the dichotomized variables based on knee extension, ankle dorsiflexion as defined, from any of the valid assessments for each study participant. If any of the variables from any of the valid assessments indicated failure, the variable takes on the value 1, if they did not, it takes on the value 0. Valid assessments were defined as any assessment between the day of surgery and end of follow-up, that does not follow a previous assessment where failure was detected.

Analyses

Baseline data will be summarized stratified by surgical group. N, Mean, Standard Deviation, Minimum and Maximum will summarize continuous variables, whereas number and percent will summarize categorical efficacy variables. ROM outcome data will be described using spagehetti plots with added group means and survival outcomes described using failure curves. Outcome data summaries will be stratified by subdiagnosis groups, USCP and BSCP, for the purpose of exploratory analysis of differences in surgical results.

Analyses of continuous primary outcomes, knee flexion and ankle dorsiflexion, will be performed using linear mixed effects modelling as described below. Differences in mean development over follow-up time will be tested using a one tailed likelihood ratio test at the 5% significance level.

Secondary endpoints will be evaluated using methods of survival analysis, such as Kaplan Meier estimates of failure curves together with confidence intervals and Cox regression analyses, as described below, will be tested using a one tailed likelihood ratio test at the 5% significance level.

Variation in differences between surgical groups comparing USCP and BSCP subgroups, will be explored in an interaction analysis using linear mixed effects regression analysis, also defined below.