Development and validation of the Charcot-Marie-Tooth Disease Infant Scale

Abstract

Many genetic subtypes of Charcot-Marie-Tooth disease (CMT) show signs of symptomatic disease during the earliest years of life. This might be the ideal time to intervene before progression of clinical sequelae due to demyelination and axonal loss. In the absence of disease-specific clinical trial outcome measures for CMT during infancy and early childhood the aim of this study was to develop and validate a functional measure of disease severity, known as the Charcot-Marie-Tooth disease Infant Scale (CMTInfS). Development projects involved identification of a preliminary pool of 31 items representing the range of disability in affected patients aged 0–4 years from a systematic review of the literature, peer review by 12 expert clinicians and researchers in the field, design of a scoring algorithm and pilot testing in 22 participants. Subsequently, a series of validation projects were conducted based on 128 assessments of: 26 confirmed cases of inherited neuropathy (17 CMT1A, one CMT1B, one CMT1D, one CMT2C, one CMT2S, two CMT4C, one CMTX3, one Riboflavin Transporter Deficiency Type 2, and one unidentified mutation); seven ‘at risk’ cases and 95 unaffected healthy controls recruited through the NIH-funded Inherited Neuropathies Consortium. Validation projects included: Item, Factor and Rasch analysis, intra- and inter-rater reliability, discriminant ability and convergent validity with the CMT Pediatric Scale (CMTPedS) for children aged 3–4 years. Development and validation projects produced a psychometrically robust 15-item scale. Rasch analysis supported the viability of the CMTInfS as a unidimensional measure of disease severity and showed good overall model fit, no evidence of misfitting items or persons and was well targeted for affected children. The CMTInfS demonstrated high intra-rater reliability [intraclass correlation coefficient (ICC)_3,1 0.999, 95% confidence interval 0.996–1.000) and inter-rater reliability (ICC_2,1 0.997, 95% confidence interval 0.992–0.999). The CMTInfS was able to discriminate between the CMT group and controls (P = 0.006), and convergent validity demonstrated good agreement between CMTInfS and CMTPedS scores (r = 0.76, P = 0.01). The final version of the CMTInfS requires 20 min to administer and is a reliable and sensitive functional outcome measure for early onset CMT and related neuropathies.

Introduction

Charcot-Marie-Tooth disease (CMT) is the most common inherited peripheral neuropathy. Over 90 causative genes for CMT have been discovered to date. CMT affects muscle strength and sensation of the lower and upper limbs, causing problems with gross and fine motor function, balance and coordination (Reilly et al., 2011). Symptom onset can occur at any age depending on the causative gene, with many subtypes causing symptoms in the earliest years of life. Infants with CMT may present with generalized hypotonia, weakness, foot deformity or gross motor and fine motor delay (Baets et al., 2011). The natural course of the disease results in progression of symptoms into adulthood at various rates, depending on the subtype of CMT. Several promising drug candidates are under development in murine models and may lead to human trials (Das et al., 2015; Zhao et al., 2017).

Multicentre clinical trials of pharmaceutical, rehabilitative and surgical interventions have been primarily pursuing treatment for adults with CMT (Shy et al., 2005). However, it is widely recognized that the most promising age to introduce disease-modifying therapies is during infancy and early childhood. There have been several randomized controlled trials of interventions for children over the age of 3 years with CMT including the evaluation of ascorbic acid, exercise and stretching (Burns et al., 2009, 2017; Rose et al., 2010). In many types of CMT, demyelination and axonal loss occurs during early childhood and intervening in the first years of life may provide the greatest potential to restore myelination and prevent axonal loss (Yiu et al., 2008). Although trials have yet to be conducted at this age, with appropriate outcome measures, there is an opportunity to halt the course of disease before the onset of long-term disability, deformity and pain.

Carefully validated, sensitive and reliable functional outcome measures to assess the effectiveness of interventions are vital for the success of clinical trials (US Food and Drug Administration, 2017). Well-validated outcome measures have been developed and implemented in clinical trials for older children and adults with CMT. The Rasch-modified CMT Neuropathy Score (CMTNSv2-R) (Sadjadi et al., 2014) for adults with CMT, and the CMT Pediatric Scale (CMTPedS) (Burns et al., 2012) for children and adolescents aged 3 to 20 years provide a reliable and valid approach to carefully measure disease progression and treatment effect. However, a recent systematic review has concluded that there are no disease-specific outcome measures validated for patients aged 3 years or younger (Mandarakas et al., 2018). In this very young population, inclusive of both infants and young children, it is challenging to design a functional outcome measure for a progressive disease; as improvement of motor skills due to growth and development must be considered in relation to disease progression, and items need to be more observable than performance-based. For example, the CMTPedS includes 11 performance-based items of strength, dexterity, sensation, gait, balance, power and endurance that rely on the child’s ability to follow standardized instructions, which is inappropriate for the assessment of infants and young children (Burns et al., 2012).

There is a variety of validation techniques to develop psychometrically robust disease-specific functional outcome measures. Classical Test Theory is often used to validate outcome measures in neurology (Hobart, 2003). Classical Test Theory, such as factor analysis, focuses on developing outcome measures based on correlations among items and each item’s ability to equally measure the same underpinning dimension. Rasch analysis, a relatively new methodology that forms part of Item Response Theory, has been used in the construction of both the CMTPedS and CMTNSv2-R. The shift towards using Item Response Theory, allows for a theoretical Rasch-built model that can estimate item difficulty and calculate the probability of a patient’s score for an item, given their performance to other items in the scale. Removing items or changing the scoring format, identifies the ideal minimum set of items that best represents the patient cohort and places them on a linear interval scale to measure each individual. The aim of this study was to develop and validate a Rasch-built functional outcome measure specifically for infants and young children with CMT, known as the Charcot-Marie-Tooth disease Infant Scale (CMTInfS).

Materials and methods

We used a range of Classical Test Theory validation techniques as well as Rasch analysis to build a clinical outcome assessment of disease severity for use in clinical practice, natural history studies and in preparation for multicentre clinical trials.

Ethics approval from all institutions, for all projects, and written informed assent/consent from all children and their families, was obtained. The protocol was approved by the ethics board of the NIH Rare Diseases Clinical Research Network (Protocol INC6611) and monitored by institutional ethics review boards at each hospital/University.

Development of the CMTInfS

Generation of the item pool

A pool of items was generated, capturing gross and fine motor development and sensation: lower limb involvement, hand dexterity and coordination, strength, balance and mobility. The item pool included measures of motor and sensory function and were selected based on disease-specificity, functional relevance, reliability/validity, responsiveness to change, duration and age-appropriateness. The item pool was reduced based on safety during administration, patient tolerance and comfort, expense, validity, reliability and sensitivity to different stages of growth and development. Items were identified from a systematic literature review of potential outcome measures for infants and young children with neuromuscular disorders that was conducted to identify items applicable to the CMT population (Mandarakas et al., 2018). There were no CMT-specific outcome measures identified. Most outcome measures retrieved from the literature were developed for spinal muscular atrophy (The Children’s Hospital of Philadelphia Infant Test of Neuromuscular Disorders, Hammersmith functional motor scale–expanded version, Test of Infant Motor Performance and the Upper Limb Module), with likely ceiling effects for CMT due to differences in their weakness profile. However, several of the items for spinal muscular atrophy assessing distal strength and gross- and fine-motor function were suitable for inclusion in the CMTInfS, such as standing, squat and recover, jumping, climbing stairs and running (gross motor skills) and pointing with a finger, drawing, removing a lid from a container and hand grip strength (fine motor skills).

To avoid ceiling effects to improve responsiveness, it was necessary to include more difficult items to assess balance and coordination in the CMTInfS. More difficult items were drawn from developmental screening tools [Bayley Scales-3rd Edition (Albers and Grieve, 2007), Brief Assessment of Motor Function (Cintas et al., 2011) and Peabody Developmental Motor Scales-2 (Folio and Fewell, 1984)], as well as outcome measures validated for other neurological conditions including: the Gross Motor Function Measure for cerebral palsy (Michaelis, 2013) and both the North Star Ambulatory Assessment (Mazzone et al., 2009) and Performance of the Upper Limb module (Mayhew et al., 2013) for Duchenne Muscular Dystrophy, to contribute to the CMTInfS. We also contacted experts in the field to identify additional items.

Peer review

Peer review to ensure quality, suitability and coverage of potential items was conducted by 12 expert clinicians and researchers from Australia, Italy, UK and USA at the Inherited Neuropathies Consortium Investigator Meeting in Miami 2016, and with follow-up video conferences and site visits. Based on expert opinion, items with insufficient face/content validity were modified or removed to form the ‘preliminary’ version of the CMTInfS containing 31 items (Supplementary Table 1).

Pilot testing

The 31 items were pilot-tested by two master clinical evaluators (M.R.M. and O.S.) with 22 children in Australia, Thailand and the USA to check for administration problems, item instructions, order and duration. The 31-item preliminary CMTInfS could be completed in 40–60 min and was well tolerated by infants and young children.

Initial scoring procedure

The preliminary CMTInfS was divided into three parts: patient profile, physical examination and functional items (Supplementary Fig. 1). The 31 functional items generated a total raw score of 0 to 85. Items 1–16 evaluated predominantly lower limb and truncal function while items 17–31 evaluated upper limb function. Items were ordered according to typical developmental milestones. Scoring ranged between 0 and 3 for most items, with some scored 0 to 2 (‘head lag’, ‘run’, ‘wrist extension’ and ‘point’), or 0 to 1 (‘unscrew lid’ and ‘button’). Higher scores indicated increased disease severity, consistent with the scoring system in the CMTPedS and CMTNSv2-R.

Evaluators were encouraged to assess items in order, using standardized equipment and toys to engage the patient. A maximum of three ‘wilful’ attempts for each item was permitted to minimize effects of fatigue and the best score for each item was recorded. Behaviour ratings were collected to assess each participant’s tolerance of the CMTInfS and to possibly explain unexpected scores. The evaluator’s overall perception of the behaviour was recorded using the Brazelton Neonatal Behavioral Assessment Scale (Brazelton and Nugent, 1995) and a 0–100 mm visual analogue scale (VAS) (Supplementary Fig. 1). The VAS provided an overall measure of how challenging or cooperative children were during assessments, with 0 representing ‘excellent’ behaviour and 100 indicative of ‘very challenging’ behaviour.

Multi-centre data collection

Infants and young children aged 0–4 years (<60 months) were eligible for inclusion. Two patient groups were recruited: Group 1 (confirmed CMT) comprising children with genetically confirmed CMT, or those with clinical and electrophysiological evidence of CMT; or Group 2 (‘at risk’ CMT): children with a family history of CMT in a first degree relative and early clinical features of the disease, without results from genetic testing or confirmatory electrophysiology. Clinical features of CMT included: muscle hypotonia; distal wasting and weakness; gait difficulty and foot drop; absence of deep tendon reflexes, particularly at the ankle; foot deformity (cavovarus or planovalgus); sensory loss (in older children). Abnormal electrophysiological findings were decreased nerve conduction velocities (<38 m/s) and/or reduced compound muscle action potentials. The status of genetic/electrophysiological testing for confirmation of a diagnosis was monitored for participants in Group 2 (at risk CMT) and confirmed cases shifted to Group 1 (confirmed CMT) for analysis. In addition, age- and gender-matched healthy participants were recruited for Group 3 (controls) to generate normative reference values to calculate the final scoring algorithm based on z-scores. Group 3 comprised infants and young children without CMT, neuromuscular disorder or any other medical condition influencing typical growth and development. For every month of age, at least one boy and one girl were assessed in Group 3. We especially targeted the youngest (6–12 months) and oldest age groups (>36 months) to include more participants as they were thought to have the most amount of variability in development. Oversampling of children in the 13–18 month age group was also undertaken, as this is when the milestone of walking is typically achieved, creating greater variability in gross motor function. The recruitment strategy for the control group was to sufficiently represent the gender and age distribution of the healthy target population.

Clinicians from the USA, UK, Italy, Thailand and Australia were trained at a face-to-face workshop, and using the ‘CMTInfS equipment and training resource kit’ and associated video resources. From July 2016 to December 2017, the 31-item preliminary CMTInfS was prospectively administered across seven sites (University of Sydney/Sydney Children’s Hospitals Network, Australia; Royal Children’s Hospital, Melbourne, Australia; University of Iowa Hospitals and Clinics, USA; University of Rochester, USA; Mahidol University/Siriraj Hospital, Thailand; C Besta Neurological Institute, Italy; Children’s Hospital of Philadelphia, USA).

Statistical analysis

A series of internal and external validation analyses were conducted in SPSS v22.0 (IBM SPSS Statistics for Windows, Armonk, NY, USA) and RUMM2030 (RUMM Laboratory Pty Ltd, WA, Australia, 2010).

Item and factor analysis

An analysis of the preliminary 31-item CMTInfS was completed to identify high (r > 0.80) or low (r < 0.30) inter-item correlations to refine the item pool. Internal consistency was assessed using a Cronbach’s coefficient alpha. Remaining items were tested for suitability to undergo principal component analysis (PCA) using the Kaiser-Meyer-Olkin value > 0.80 (Kaiser, 1974) and the Bartlett’s Test of Sphericity (Bartlett, 1950) with a value P < 0.05.

Rasch analysis

Rasch modelling was conducted to test how well the observed data fit the expectations of the measurement model. Rasch analysis is an iterative process to achieve acceptable fit. The Partial Credit Model was used (Cappelleri et al., 2014) and three overall fit statistics were considered, relating to the item–person interaction and the item–trait interaction statistics. The acceptable range for standardized fit statistics was set at 1.5 (Shea et al., 2009). Item-trait interaction was reported as a chi-square value. Bonferroni corrections were applied to adjust the chi-squared P-value. An estimate of the internal consistency reliability of the scale was also completed, based on the Person Separation Index (PSI). Disordered thresholds were examined, and the rating scale was adjusted to resolve issues where there were too many response options. When necessary, items were removed to resolve issues with uniform and non-uniform differential item functioning. Finally, a PCA of the residuals was undertaken to detect signs of multidimensionality. Unidimensionality was confirmed if <5% of persons had significant differences in scores on each of the subsets identified.

Intra- and inter-rater reliability

Intra-rater and inter-rater reliability was examined for the resultant item pool determined by Item, Factor and Rasch analysis. For intra-rater reliability, 21 CMTInfS assessments from patients with CMT and healthy controls were scored live by the lead investigator (M.R.M.), which were also captured with video. Videos were coded (names and identifying participant information was removed) and randomly ordered by an independent research assistant. After a period of 6 weeks to 12 months, assessments were rescored by the same examiner from the videos. Kappa was calculated to assess agreement between individual raw item scores and an intraclass correlation coefficient (ICC_3,1) assessed the agreement of total scores between testing occasions. To assess inter-rater reliability, agreement between eight trained evaluators (four physiotherapists, one paediatric neurologist, one paediatrician, one occupational therapist, one podiatrist) was examined using video recordings for 10 CMTInfS assessments. An ICC_2,1 was calculated for total scores across multiple evaluators.

Discriminant ability

Total scores from the resultant Item, Factor and Rasch analyses were transformed to z-scores based on the age- and gender-matched normative reference values from the control group, in accordance with CMTPedS scoring methodology (Burns et al., 2012). Z-scores represent the number of standard deviations participants score away from ‘normal’ function (defined as 0). Z-scores >0 indicate increasing levels of disease severity. To assess discriminant ability, z-scores from patients in the ‘CMT group’ were compared to those of healthy controls using an Independent samples t-test.

Convergent validity

Children aged 3 and 4 years were assessed with both the CMTInfS and the CMTPedS to determine convergent validity. Scores from the two clinical outcome assessments were evaluated using a Bland-Altman plot to establish distribution within Limits of Agreement values in addition to examining the Pearsons’s correlation coefficient.

Longitudinal responsiveness

A subset of infants and children with CMT were reassessed after 1 year with the CMTInfS, and the median [interquartile range (IQR)] change was calculated for z-scores.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Results

Participants

A total of 128 children aged 0–4 years (26 confirmed CMT, seven ‘at risk’ of CMT and 95 healthy controls) were assessed with the CMTInfS (participant characteristics detailed in Table 1 ). At the time of analysis, three participants in the at-risk CMT group received confirmatory genetic testing (one CMT1A, one CMT4C, in addition to one patient with Riboflavin Transporter Deficiency type 2, RTD2) (recruitment numbers for each group at each stage of analysis are detailed in Fig. 1). The control group included boys and girls assessed in 6-month blocks of ages from 6 to 36 months: 6–11 months n = 15 (67% female), 12–17 months n = 17 (59% female), 18–23 months n = 14 (57% female), 24–29 months n = 14 (43% female), 30–35 months n = 14 (36% female), and a group of 20 children aged 36–58 months (40% female) for patient follow-up comparisons. The control group was recruited in Australia n = 71 (74%), Thailand n = 27 (28%) and the USA n = 1 (1%). CMTInfS scores in the control group were comparable between countries (one-way ANOVA P = 0.18).

Behaviour during CMTInfS assessments was generally good between groups [CMT group: mean VAS 12.0, standard deviation (SD) 19.0 mm; at-risk CMT group: mean VAS 30.0, SD 29.0 mm; control: mean VAS 9.4, SD 15.9 mm]. Brazelton state was scored as a 5 (‘completely awake, strong motor activity’) for 24/26 (92%) of patients with CMT, 6/7 (86%) of the at-risk CMT group, and 93/95 (98%) of controls.

Item analysis

High inter-item correlations in the preliminary CMTInfS, indicative of potential redundancy, were observed between the following items: ‘stand’, ‘cruise’, ‘walk’, ‘jump’, ‘hop’, ‘step’, ‘scissors’ and ‘pincer grasp’ (r > 0.80, Supplementary Table 1). Items that did not correlate substantially with any other items were ‘head control’, ‘head lag’, ‘lift leg in supine’, ‘digital extension’ and ‘wrist extension’ items (r < 0.30, Supplementary Table 1). These highly and poorly intercorrelated items were removed from the CMTInfS. Item analysis of the resultant 18 items showed a full range of scoring from a low of 0 (able/normal) to a high of 3 (or maximum 1 or 2 for items with different scoring structures: ‘run’, ‘point’, ‘unscrew lid’ and ‘button’; indicating unable/severe). Mean scores ranged from a low of 0.05/3 (easiest: ‘sit’) to a high of 0.77/1 (hardest: ‘button’). In the inter-item correlation matrix for the 18 items, no items were too highly (r > 0.80) inter-correlated and each item correlated with at least one other with r > 0.30. Average inter-item correlation was 0.42 (range − 0.03–0.78). The Corrected Item-Total Correlation (the degree to which each item correlates with the total score) ranged from 0.19 (‘prop’) to 0.83 (‘build tower’). Internal consistency, or the degree to which the 18 items group together, was high with a Cronbach’s coefficient alpha value of 0.93. However, such a high Cronbach’s coefficient alpha value indicated the scale required further removal of items to address redundancy (DeVellis, 2003).

Factor analysis

The remaining 18 items were subjected to PCA. The Kaiser-Meyer-Olkin value (0.89) and the Bartlett’s Test of Sphericity (P < 0.001) supported the factorability of the correlation matrix. PCA revealed the presence of three components with eigenvalues exceeding 1, explaining 47%, 14% and 7% of the variance respectively for a total of 69%. The scree plot (Supplementary Fig. 2) demonstrated a clear break between the second and third components, supporting the retention of two components for further investigation. This decision was further supported by the results of parallel analysis, which showed two components with eigenvalues exceeding the corresponding criterion values for a randomly generated data matrix of the same size (18 items × 128 cases, Supplementary Table 2). The two-component solution explained 62% of the variance (Component 1 contributing 47%, Component 2 contributing 14%). To aid in the interpretation of these two components, Oblimin rotation was performed (Supplementary Table 3) (Tabachnick and Fidell, 2007). The Oblimin rotated solution showed both components having a number of strong cross loadings. One item (‘palmar grasp’) did not explain much variance, with a communalities extraction value of 0.26.

Interpreting the two components indicated that the 11 items in Component 1 represent the most difficult items (difficult gross motor skills beyond typical age of walking and the most difficult bimanual fine motor tasks) and Component 2 is representative of the seven easiest items (including gross motor skills developed before the age of walking and single-handed fine motor skills including reach and palmar grasp). When forcing a single-factor solution, 15 of 18 items loaded onto the first factor (>0.5), supporting the unidimensionality of the scale and a 1-factor solution to be most appropriate (Supplementary Table 4). A decision on whether to delete ‘reach’, ‘prop’ and ‘sit’ items was made during Rasch analysis.

Rasch analysis

Item reduction

Dimensionality testing indicated that 5/128 (4%) children had statistically different scores (P < 0.05) on the two sets of items (upper and lower limb items) identified from PCA of the residuals, suggesting the scale was initially unidimensional. Items re-scored during the Rasch analysis included ‘prop’, ‘squat’ and ‘sit’, due to misfit and disordered thresholds. Additional analyses were performed to guide retention of items discriminating between the CMT and control groups. Commencing with the list of items to refine first, whilst maintaining important discriminating items, items were either rescored, one at a time to improve model fit (increase chi-square probability and address disordered thresholds), or removed. With the rescoring of seven items and the reduction of three items (‘prop’, ‘grasp marker’ and ‘reach’) (Supplementary Table 5), the 15-item revised version of the CMTInfS showed that only 4/128 (3%) [95% confidence intervals (CI) −0.7–6.9%] children scored significantly different results on the subsets identified.

Rasch modelling

There was good overall model fit (chi-square probability 0.15, non-significant P-value with Bonferroni correction 0.05/15 = 0.003) for the 15-item CMTInfS. There was no evidence of misfitting items (fit residual mean: −0.38, SD 0.77) with no fit residuals >1.5 and no significant chi-square probability values. There was no person misfit (mean: −0.19, SD 0.27) with no fit residuals >1.5. There was acceptable person separation reliability to differentiate groups of patients (Cronbach’s alpha 0.93). There were no items showing uniform/non-uniform differential item functioning. Residual correlations showed no evidence of serious response dependency. Two disordered thresholds were present (roll and sit items); however, these had no effect on overall model fit and did not cause item misfit. The Person-Item distribution shows the CMTInfS is well-targeted.

The item map showed a good spread for the 15-item scale (Supplementary Fig. 3). Items spread across the full range of scores for this sample, with no gaps where there were insufficient items to assess specific levels of disease severity. Tearing paper was the most difficult item (i.e. commonly severely affected) and sitting independently was the easiest item (i.e. rarely severely affected). There was no clustering at the low or high ends of the distribution (floor or ceiling effects). The final version of the CMTInfS is shown in Fig. 2 and takes ∼20 min to administer.

Reliability

The intra-rater reliability analyses involved 21 infants and young children (two CMT1A, 19 controls) aged 6–58 months (mean age: 34.1 months, SD 16.6). Good to excellent intra-rater agreement for each of the 15 items was demonstrated, with Kappa scores ranging from 0.57 (‘palmar grasp’) to 1.00 (eight gross and fine motor items) (Supplementary Table 6). The mean Kappa value was 0.88 for individual raw item scores. Total CMTInfS scores were calculated for 12 participants with complete videos available (mean age 31.7 months, SD 17.7, range 8–58) and produced ‘excellent’ agreement (ICC_3,1 0.999, 95% CI_3,1 0.996–1.000). Only two scores differed by one point, falling outside the Limits of Agreement on the Bland-Altman Plot (Fig. 3). For inter-rater reliability, videos of 10 CMTInfS assessments (four CMT: three CMT1A and one CMT4C, and six controls; mean age: 29.7 months, SD 17.9, range 5–58) were scored. The total CMTInfS z-scores had ‘excellent’ inter-rater reliability (ICC_2,1 0.997) with narrow 95% CIs (95% CI 0.992–0.999).

Discriminant ability

Z-scores were calculated based on the age- and gender-matched normative reference values from the control group (Table 2). Infants and children in the ‘CMT group’ (n = 26) scored significantly higher (more severe) on the CMTInfS (mean z-score 1.6, SD 2.2, range − 0.2–7.7) compared to controls (n = 95) (mean z-score 0.0, SD 1.0, range − 2.0–3.2) (independent samples t-test, P = 0.001) (Fig. 4). CMTInfS scores also discriminated between patients with CMT1A (n = 17, mean age: 37 months, SD 16.5, 41% female, mean CMTInfS z-score: 0.8, SD 1.3, range − 0.2–3.9) compared to the control group (n = 95, mean age: 25.3 months, SD 13.6, 49.5% female, mean CMTInfS z-score 0.0, SD 1.0, range − 2.0–3.2, independent samples t-test, P = 0.006). Patients with the most severe neuropathy subtypes (CMT1D, CMT2S and RTD2) had the highest CMTInfS z-scores (Fig. 4).

Convergent validity

Ten children (mean age 46.6 months, SD 7.7, 50% female) comprising seven CMT1A, one CMT1D, one CMT4C and one at-risk of CMT case were assessed using both the CMTInfS and the CMTPedS. There was good agreement between scores (Pearson’s r = 0.76, P = 0.01), with no scores lying outside limits of agreement values in the Bland-Altman plot (Supplementary Fig. 4). For CMTInfS z-scores, severity was also categorized as mild (<1.0), moderate (1.0–2.0) and severe (>2.0) and for CMTPedS scores as mild (0–14), moderate (15–29) and severe (30–44) (Table 3) (Cornett et al., 2017). For the two discrepancies (Table 3), children had more severe scores on the CMTInfS than the CMTPedS.

Longitudinal responsiveness

Four patients with CMT1A (mean age 20.5 months, SD 13.5, range 8–35 months) were reassessed with the CMTInfS after 1 year (mean time of follow-up: 12.8 months, SD 0.5, range 12–13 months). Median (IQR) change was 0.29 (IQR 0.17–1.89) indicating an overall progression of disease severity. Two patients worsened (percentage change 276–343%) and two remained relatively stable (percentage change in score <+35%) (Fig. 5).

Discussion

The CMTInfS is a psychometrically robust 15-item functional outcome measure suitable for use in clinical practice, natural history studies and in preparation for multicentre clinical trials for affected patients at the earliest stage of their disease. The Rasch-built CMTInfS can distinguish between patients and controls, and early observations suggest that the scale is sensitive to genetic subtype, with scores corresponding to known phenotypes and expected levels of severity.

The CMTInfS scoring algorthim is based on normative reference values, so that infants and young children at all stages of development can be compared on the same scale of disease severity. One of the limitations identified in our systematic review of existing outcome measures for infants with neuromuscular disorders was that the methodology for item scoring did not always account for growth and development (Mandarakas et al., 2018). The use of z-scores enables clear interpretation of disease severity within and between children over time. A ‘moderate’ disease severity rating (z-score 1.0–2.0) translates to the lowest 16% of function. A ‘severe’ disease severity rating (z-score > 2.0) means that children are functioning in the lowest 3% of scores. The CMTInfS was able to capture all levels of disease severity, including the most severely affected children with z-scores as high as 7.7. Further, none of the outcome measures included in the systematic review were validated with Rasch analysis, a relatively new methodology that forms part of Item Response Theory, used in the construction of both the CMTPedS and CMTNSv2-R (Mandarakas et al., 2018). Using Rasch analysis to validate the CMTInfS ensured unidimensionality with no evidence of differential item functioning or item misfit, providing the best possible scoring format to produce a valid and sensitive outcome measure.

Investigating the transition from the CMTInfS to the CMTPedS is our next step to enable tracking of the disease course over the entire lifespan. While preliminary results support convergent validity, with good agreement between CMTInfS and CMTPedS in children able to complete both assessments, more data are needed to enable the transition between scores in longitudinal studies. Of the 10 children assessed with both the CMTInfS and CMTPedS, one child (aged 46 months) with CMT1D showed the largest difference in scores and was classified as ‘severe’ on the CMTInfS (z-score 7.7) and only ‘moderate’ on the CMTPedS (27/44). Clinically, this patient presented with a severe disease phenotype, being unable to walk or stand without bilateral ankle-foot orthoses and had impaired sensation and hand function. As the child was in the upper most age limit of the CMTInfS, any score that deviated from 0 (able) resulted in a more severe z-score. For the CMTPedS, however, being unable to attempt an item, in some instances, resulted in only ‘mild’ or ‘moderate’ severity ratings for individual items, leading to a ‘moderate’ severity rating overall. The CMTPedS involves some difficult items that require advanced levels of comprehension, so high variability is seen in children <4 years of age. Therefore, for children aged 3 years, the CMTInfS may be more sensitive to disease severity and clinical impression.

This study was conducted as part of the Inherited Neuropathies Consortium, an international group of CMT specialists collecting longitudinal data using both the CMTPedS and CMTNSv2. Infants and young children assessed with the CMTInfS will contribute to this database of almost 10 000 patients. Training of clinical evaluators across the Inherited Neuropathies Consortium will commence using the revised ‘CMTInfS equipment and training resource kit’ (Supplementary material). A system of quality assurance is under development to ensure all evaluators receive training according to their level of experience and qualifications, including, but not limited to: face-to-face training sessions, feedback on videoed assessments, the provision of online training resources and regular follow-up meetings to discuss issues with CMTInfS implementation. An online scoring system will also be developed to simplify the generation of z-scores from raw data. This approach will ensure clinical trial-readiness for each site in preparation for therapeutic trials at the earliest stage of disease.

This study has some limitations. CMT is heterogeneous in nature, with different subtypes progressing at substantially different rates. Whilst the CMTInfS responded as expected, with CMT1D, CMT2S and RTD2 scoring highest and the mildest type, CMT1A, scoring lowest, a larger sample size of patients, especially with rarer genetic subtypes is required. Longitudinal data are crucial to assess scale responsiveness and power clinical trials. Although our preliminary prospective data for four patients with CMT1A showed possible overall progression of disease severity, longer-term follow-up for more patients is necessary to make any definitive statements about progression. A longitudinal study will be actively pursued in a larger group of CMT patients across the Inherited Neuropathies Consortium.

To conclude, the final version of the CMTInfS requires 20 min to administer and is a reliable and sensitive 15-item functional outcome measure for early onset CMT and related neuropathies. The CMTInfS will be implemented in natural history studies to understand the rate of progression at the earliest stages of disease in preparation for therapeutic trials of early interventions.

Acknowledgements

We would like to thank the children and their families for their participation in this study. Thank you also to those who assisted in the organisation and assessment of patients: Kayla Cornett, PhD (The Children’s Hospital at Westmead, Sydney Australia), Apirada Thongsing, MD and Sivaporn Limpaninlachat, PT (Siriraj Hospital Mahidol University, Bangkok Thailand), Shawna Feely, LCGC (University of Iowa Hospitals and Clinics) and Carly Siskind, LCGC (Stanford University) and to Kathy Priest for her assistance in deidentifying reliability data (The Children’s Hospital at Westmead, Sydney Australia). We are thankful for the statistical guidance, review and expertise provided by Associate Professor Julie Pallant for the Rasch analysis and to Richard Finkel, MD (Nemours Children’s Hospital, Orlando, USA), Allan Glanzman, PT, DPT, PCS (Children’s Hospital of Philadelphia) and Anne Connoley, MD (Washington University, St Louis USA) for their assistance with scale development. We are grateful for the support of site coinvestigators: Mary M. Reilly, MD (MRC Centre for Neuromuscular Diseases, Queen Square, London, UK), Matilde Laura (UCL Institute of Neurology, Queen Square, London, UK), Francesco Muntoni, MD and Trupti Bhandari, PT (Great Ormond Street Hospital, London, UK), Sylvia Ounpuu, PT, Allison Fullam PT and Kendra Bow, PT (Connecticut Children’s Medical Center).

Funding

This project was supported by the Inherited Neuropathies Consortium (INC), part of the Rare Diseases Clinical Research Network, and was funded by the National Institutes of Health (#U54NS065712 supported by National Institutes of Neurological Diseases and Stroke and Office of Rare Diseases). The INC also receives support from the American Muscular Dystrophy Association and the Charcot Marie Tooth Association (CMTA). This project was also supported by the University of Sydney South East Asia Centre Cluster Research Grant (2015).

Competing interests

J.B received support from Charcot-Marie Tooth Association of Australia and CharcotMarie Tooth Association (USA). M.E.S. received support from the Muscular Dystrophy Association, Charcot Marie Tooth Association and Inflects Pharmaceuticals and is a consultant for Alnylam Pharmaceuticals, Flex Pharma and Accelerant Pharmaceuticals. D.P. received support from the Charcot Marie Tooth Association. K.J.R. has consultancy agreements with Biogen Idec, IONIS Pharmaceuticals, Hoffman-La-Roche and ATOM International. E.M.Y. is supported by a NHMRC Early Career Fellowship. M.P.M was supported by a grant from the Thyne Reid Foundation. K.E. receives consulting fees from Acceleron Pharmaceuticals.

Abbreviations

Abbreviations
 
• CMT

  Charcot-Marie-Tooth disease

 
• CMTInfS

  Charcot-Marie-Tooth disease Infant Scale

 
• CMTPedS

  Charcot-Marie-Tooth disease Pediatric Scale

 
• CMTNSv2-R

  Rasch-modified CMT Neuropathy Score 2nd version

 
• ICC

  intraclass correlation coefficient

 
• PCA

  principal component analysis

 
• VAS

  visual analogue scale

References

Author notes