RASCH ANALYSIS OF CLINICAL OUTCOME MEASURES IN SPINAL MUSCULAR ATROPHY

Abstract

Introduction

Trial design for SMA depends on meaningful rating scales to assess outcomes. In this study Rasch methodology was applied to 9 motor scales in spinal muscular atrophy (SMA).

Methods

Data from all 3 SMA types were provided by research groups for 9 commonly used scales. Rasch methodology assessed the ordering of response option thresholds, tests of fit, spread of item locations, residual correlations, and person separation index.

Results

Each scale had good reliability. However, several issues impacting scale validity were identified, including the extent that items defined clinically meaningful constructs and how well each scale measured performance across the SMA spectrum.

Conclusions

The sensitivity and potential utility of each SMA scale as outcome measures for trials could be improved by establishing clear definitions of what is measured, reconsidering items that misfit and items whose response categories have reversed thresholds, and adding new items at the extremes of scale ranges.

Spinal muscular atrophy (SMA) is a rare¹ inherited neuromuscular condition characterized by degeneration of spinal and bulbar anterior horn cells with resulting muscle weakness and atrophy²; it is commonly classified into 1 of 3 categories.³ Over the past 15 years increased understanding of the molecular genetics and pathophysiology of SMA has provided a platform for therapeutics development and new promising clinical trials^4–6 whose efficacy is ultimately judged by outcome measures. Functional rating scales are being used increasingly in this important role,^7–9 and it is therefore essential that they provide clinically meaningful and scientifically robust data. These issues highlight the key role for psychometric methodologies in scale development and testing.¹⁰ Many rating scales have been used as outcome measures in SMA natural history studies and clinical trials. Collectively the scales are designed to capture disease progression due to muscle weakness in the upper and lower extremities and the trunk that are the clinical hallmarks of the disease. In capturing aspects of underlying weakness, the various movement and functional measures have been shown to document change over time¹¹ and are further expected to reflect any changes that might result from treatment interventions that may range from slowing of disease progression to functional improvement.

The most commonly used motor performance scales include: Hammersmith Functional Motor Scale (HMFS),^12,13 the Expanded HFMS (HFMSE),¹⁴ Modified Hammersmith Functional Motor Scale Extend (MHMS-Extend),^15,16 Motor Function Measure (MFM),¹⁷ Gross Motor Function Measure (GMFM),^18,19 North Star Ambulatory Assessment for SMA (NSAA-SMA),^20,21 Egen Klassifikation Scale v.2 (EK2),²² Children’s Hospital of Philadelphia Test of Neuromuscular Disorders (CHOP-INTEND),^23,24 and Test of Infant Motor Performance Screening Items (TIMPSI).²⁵ A summary of the scales is listed in Table 1. Although these scales have undergone varying degrees of traditional psychometric testing,^{12,13,15,17,20–23,26–29} they have not been evaluated with Rasch measurement methods in SMA.³⁰ The Rasch Measurement Model (RMM) is a mathematical model that is used to verify whether or not adding scores from ordinal scale measurements (such as the Hammersmith Functional Motor Scale) is justified by the data and also allows conversion of ordinal level data to interval level data, allowing subjects to be placed along an interval continuum of ability. Rasch analysis has the potential to test for and provide evidence to help advance the scientific merit of commonly used outcome measures. It has been applied successfully to outcome measures in several neurological disorders, including cervical dystonia, Friedreich ataxia, and multiple sclerosis.^31–33 The analysis itself is based on the requirements that: (1) more able individuals will perform better on more test items; (2) fewer individuals will be able to perform well on more difficult, challenging test items; and (3) responses to items of a scale clearly represent different amounts of a single skill or attribute.

Table 1.

Summary of analyzed scales.

Scale	Abbreviation	Summary details of scale	Traditional reliability tests performed specific to use in SMA?
Children’s Hospital of Philadelphia Test of Neuromuscular Disorders23,24	CHOP-INTEND	Designed for use in young children with severe forms of NMD. 16 items observe spontaneous movements and elicited reactions. http://ebookbrowse.com/chop-intend-for-sma-type-i-manual-of-procedures-pdf-d87663908	Y
Egen Klassifikation Scale v.222	EK2	Mainly question-based scale of 17 items relating to an individual’s “own functioning” in the home.	Y
Gross Motor Function Measure18,19	GMFM	Originally developed for use in cerebral palsy. Contains 88 items and 5 dimensions of lying and rolling, sitting, crawling and kneeling, standing and walking, running and jumping. http://motorgrowth.canchild.ca/en/gmfm/resources/gmfmscoresheet.pdf	Y
Hammersmith Functional Motor Scale12,13	HFMS	Developed for use specifically in non-ambulant SMA. 20 items including lying, rolling, sitting, and into standing and stepping.	Y
Expanded Hammersmith Functional Motor Scale14	HFMSE	An expanded version of the HFMS with additional items particularly aimed at ambulant individuals (additional 13 items)	Y
Motor Function Measure17	MFM	A 32 item scale with 3 dimensions, D1 (standing and transfers), D2 (axial and proximal motor function), and D3 (distal motor function). It was designed to be suitable for all neuromuscular conditions, not just SMA. http://www.mfm-nmd.org/home.aspx	N
Modified Hammersmith Functional Motor Scale Extend15,16	MHFMS-Extend	Modified version of the HFMS, with defined scoring, a change in the testing order, and an extended module (8 additional items) which allows assessment of individuals with SMA who are non-ambulant as well as ambulant. http://smaoutcomes.org/hammersmith_manual/manual_1.html	Y
North Star Ambulatory Assessment for SMA20,21	NSAA for SMA	The original 17 item scale was designed for use in ambulant DMD and adapted for use in SMA. It includes items such as stepping on and off a box, jumping and hopping.	N
Test of Infant Motor Performance Screening Items25	TIMPSI	Designed originally for assessing infants at risk from developmental delay. This 29 item scale has been used subsequently for assessment of severely affected SMA individuals.	N (Y)

Rasch measurement methods are based on Rasch measurement theory³⁴ and are used to examine the extent to which the observed rating scale data (in this instance clinicians’ ratings on scale items) “fit” with predictions of those ratings from the Rasch model (which defines how a set of items should perform to generate reliable and valid measurements).³⁵ Effectively, the difference between expected (as predicted by the model) and observed scores indicates the degree to which valid measurement is achieved. In contrast to traditional psychometric methods, Rasch measurement methods are better able to: identify the strengths and weaknesses of rating scales⁹ while at the same time ensuring that scales fulfill currently mandated traditional psychometric criteria³⁶; provide data that are more clinically interpretable³⁷; and provide detailed evidence-based recommendations for scale improvement.¹⁰ Importantly, the use of Rasch measurement methods is supported by regulatory bodies such as the Food and Drug Administration (FDA).³⁸ Further details on this methodology are described elsewhere.^10,30,39

This article summarizes work carried out by a Task Force of the International Coordinating Committee for SMA Clinical Trials (ICC) that resulted in a summary meeting held in May 2011. The ICC and Rasch Task Force are made up of an international group of investigators who work together to address clinical trial issues in the SMA community including reaching expert consensus on the use and utility of functional rating scales measures and study endpoints suitable for use in pivotal efficacy studies.⁴⁰ The Rasch Task Force included representatives of all the study groups involved in developing the most widely used functional rating scales in SMA (see Acknowledgements). Over the past 20 years, despite the rarity of this condition these groups have amassed collectively sufficient data to facilitate a detailed psychometric analysis of each scale. The focus of this study was to use the benefits of Rasch measurement methods to gain a better understanding of the psychometric strengths and weaknesses of current SMA motor performance rating scales with a goal of making recommendations for future development and improvement for use as clinical trial outcome measures.

METHODS

Participants, Recruitment, Data Collection, and Scales

Datasets from children and adults with SMA types 1, 2, and 3 were made available from research groups representing each of the rating scales to be analyzed. Details are listed in Table 2 (additional information on patient characteristics of the scales is available in Supplementary Table S1, which is available online). A request to use these datasets was supported and organized by the ICC for SMA Clinical Trials. All data were provided in de-identified format consistent with the original data collection IRB approvals obtained by the research groups. The goal was to achieve sample sizes for each dataset above the recommended figure for item analysis (n > 300).⁴¹ Nine motor performance rating scales were selected for analysis (Table 2).

Table 2.

Sample size of dataset made available for each scale.

Abbreviation	Study Group	Sample size (person measurements)	SMA Type
CHOP-INTEND	PNCRN*	132	1, 2
EK2	Danish Muscle Group, Italian SMA Group, UK Smartnet, PNCRN	154	1, 2, 3
GMFM	AmSMART PNCRN	554	1, 2, 3
HFMS	Italian SMA Group, Hammersmith Clinic/SMAF	390	2, non-ambulant 3
HFMSE	PNCRN	579	1, 2, 3
MFM	MFM Study Group	143	1, 2, 3
MHFMS	Project Cure, SMAF- BforSMA Study	401	2, 3
NSAA for SMA	Smartnet – Part of North Star Clinical Network, UK	48	3
TIMPSI	Project Cure	83	1

^*PNCRN, Pediatric Neuromuscular Clinical Research Network for SMA.

Rasch Measurement Methods

The data were analyzed using RUMM2030,⁴² which is a Microsoft Windows-based psychometric item analysis software package. It includes the Rasch measurement algorithms and tests developed by Georg Rasch³⁵ and David Andrich, the developers of Rasch measurement theory.^30,43–46 Five key tests for reliable and valid measurement were assessed in this analysis, including fit, targeting, dependency, reliability, and stability.

Fit

The individual items in any motor performance scale must work together (i.e., fit) as a conformable set, both clinically and statistically. Otherwise, it is inappropriate to sum item responses to reach a total score and consider the total score as an overall measure of motor performance. When items do not work together (i.e. misfit) in this way, the validity of a scale is uncertain. For example an item reporting fatigue does not fit the construct of motor performance. In brief, for this analysis, evidence for item fit was based on 4 different indicators. These included ordering of item response options, ordering of item thresholds,⁴⁷ 2 statistical indicators (fit residual; χ square), and 1 graphical indicator (item characteristic curve; Item CC⁴⁸).

Targeting

Scale-to-sample targeting refers to the match between the range of motor performance measured by each of the 9 scales and the range of motor abilities in the sample of patients in each dataset. This analysis involves examination of the spread of person (individuals who performed the scale) and item locations in these 2 relative distributions. This analysis indicates how suitable the patient sample is for evaluating the scales and how suitable the scales are for measuring their respective samples. Better targeting equates to a better ability to interpret the psychometric data with confidence.⁴⁸

Dependency

Dependency refers to the degree to which the response to any of the items in each of scale are directly influenced by the response to any other item in the same scale.⁴⁹ For example, the scores for rolling to the right and left may be strongly associated. If this occurs, measurement estimates may be biased, and reliability may be elevated artificially. Rasch measurement methods determine this effect by examining residual correlations.

Reliability

Reliability refers to the extent to which scale scores are not associated with random error. This was assessed using the Person Separation Index (PSI),⁴⁴ comparable to the Cronbach α.⁵⁰

Stability

It is important that the items of each scale perform in a similar way across subgroups of children that could be compared (e.g., gender, different age ranges). The degree to which item performance remains stable across subgroups is known as differential item functioning. For this analysis, we tested SMA type, age, and study group.

RESULTS

Datasets

Scales were analyzed in datasets that included SMA Types 1, 2, and 3, except: TIMPSI (Type 1); NSAA for SMA (Type 3); CHOP-INTEND (Types 1 and 2); and HFMS, MHFMS (Types 2 and 3). The CHOP-INTEND, EK2, MFM, NSAA, and TIMPSI included datasets of n < 300, which is less than ideal.⁴¹ Therefore, the findings of the analyses should be interpreted with this in mind. However, analysis on smaller subsets can provide useful exploratory information and can direct future research and scale development.

Findings from Rasch Analysis

Given the very large amount of data produced by each of the rating scales, the data are presented as an overall summary of the findings, with additional illustrative examples of some of the key findings. Supplementary Table S2 includes a summary of the findings for the 9 rating scales. All Rasch data pertaining to the scales is available from the authors; data for each individual scale is available from the contributing research group. Readers interested in further information related to the specific findings should contact the corresponding author (J.T.K.; Chair of the ICC Rasch Task Force).

Fit

All of the 9 rating scales were found to have problems with fit. Thus, between 12% (NSAA) and 88% (GMFM) of items were found to have reversed thresholds, meaning that their response categories are not working as intended (see the Discussion section for further elaboration). All of the scales except the CHOP-INTEND showed problems with fit residuals and χ² tests. For example, approximately one-third of the HFMS items had fit residuals outside 22.5/12.5, and approximately onethird of the HFMSE items had χ² probabilities >0.001 (Supplementary Table S2). This suggests that these items do not fit ideally with the intended construct of the scale. This could be because the scale is multidimensional (i.e., not measuring 1 domain such as ambulatory function).

Targeting

The TIMPSI, CHOP-INTEND, HFMSE, and MHFMS-Extend showed adequate scale targeting to the samples under study. This indicates that the range of items covered the range of motor performance ability seen in the associated patient samples. The remaining scales were not as well targeted to their respective study samples (see Tables 1 and 2). As such, the EK2, GMFM, and HFMS are better targeted to less impacted children, and the MFM and NSAA are better targeted to more impacted children (Fig. 1). This must be considered in the context of a scale’s original target group.

FIGURE 1.

Motor Function Measure (MFM), Person-Item thresholds distribution. The upper histogram represents the sample distribution of total MFM person measures for spinal muscular atrophy (SMA) type 1 (blue), SMA type 2 (red), and SMA type 3 (Green). The lower histogram blue-striped blocks represent the sample distribution of the item thresholds of the 32-items of the MFM. The green line shows the information function. The graph shows that the distributions of item thresholds and person measures are reasonably matched. There are a few areas where there are gaps on the item threshold continuum, which impacts precision, and other places where item threshold locations are close, which indicates potential redundancy. Although the bimodal distribution indicates 2 areas of optimal measurement (at the peaks), the highest peak of the information plot is around −22.0 logits of the continuum. This suggests that the scale is probably better targeted for patients at the more impacted end of the continuum, for children with SMA Type 1 or Type 2.

Dependency

A proportion of the items, in all except 2 of the scales (EK2; see Supplementary Table S2), showed some local dependency between pairs of items. This suggests that the response to 1 or either of the items in the pair is influenced directly by the others. This ranged between 1 pair (CHOP-INTEND) to 61 pairs of items (GMFM) showing pairs of items with residual correlations >0.40. For example, the dependency shown by the residual correlation (r = 0.69) between the GMFM “walks 10 steps supported” and “walks 10 steps independently” indicates the response pattern to the latter item is influenced by the former. High levels of dependency can artificially inflate reliability.

Reliability

All scales showed adequate reliability with PSI > 0.83 within the context of local dependency issues.

Stability

All scales tested for stability except the TIMPSI were found to have differential item functioning (DIF), suggesting patterns of responses which differ systematically between different groups to differing degrees. The scales most affected were the MHFMS Extend (DIF by SMA type for 43% of items), HFMSE (DIF by age for 55% of items), and the HFMS (DIF by study group for 25% of items).

DISCUSSION

The 9 scales examined in this study are all invaluable as clinical assessment tools. However, to be considered appropriate measurement instruments for clinical trials, a series of psychometric criteria must be met. In this study, Rasch measurement methods allowed us to gain a detailed level of understanding of the measurement performance of the items of each rating scale and to move toward standards set by regulatory authorities. The analysis revealed psychometric strengths and weaknesses for all of the scales. All of the scales demonstrated adequate reliability (which may be influenced by dependency), but there were validity-related problems with the extent to which some items in each of the scales measure adequately motor performance in children with different levels of SMA. In this regard, for a proportion of the items in all of the scales, there was a problem with ordering of response categories, suggesting that for these items the scoring function does not work as intended in children with SMA.

For example, for item 2 of the GMFM (“Hands to midline”), one would expect that a child who has mild motor performance problems as measured by the overall GMFM would, probabilistically-speaking, score a 3 (“completes” task), while a child with more severe involvement would score lower. Thus, a child with moderate problems would probably score a 1 (“initiates” task) or a 2 (“partially completes” task), and a severely affected child would probably score 0 (“does not initiate” task). However, Rasch analysis showed that this was not the case for this and many other items in the GMFM, and there were similar problems in each of the other scales. This reversal of response option ordering, known as disordered thresholds, may have occurred for several reasons, including the fact that there may have been too many response options for evaluators.¹⁰ Another possibility is that there may be something specific about the way clinicians interpret and score the children using the scale categories that does not fit within the frame of reference of the scales. Thus, for the GMFM in the above example, there is a consistent problem in the way the “partially completes” and “completes” categories were scored. Although the analyses do not reveal the precise cause, this is important, as ordered thresholds are fundamentally important for establishing scale validity.^10,43,51

Reversed thresholds can also contribute to item misfit, another problematic feature identified in many of the scales. However, item misfit may have also arisen directly from the scale content. Of necessity, measuring motor performance covers a wide range of clinical presentation and muscle weakness, so that some of the scales may include items that are tapping into different constructs. For example, the HFMSE includes items that include head movement tasks (item 17, supine lifts head), upper limb tasks (e.g., item 4, 2 hands to head) and lower limb tasks (e.g., item 32, up stairs no rail). This issue, known as multidimensionality,⁵² is potentially problematic, as the different sets of items represent different types of variables with different roles.⁵³ A useful analogy is the comparison of height and weight. Although we would expect height and weight to be correlated, the combination of these measures to form a “total score” would probably be neither helpful nor meaningful. In the same way, although one would expect the items of the HFMSE to be correlated, given their relationship in SMA as affecting the whole body, combining them into a total score will confound measurement and makes interpretation difficult.⁵³ Adding together items from related but potentially different constructs could contribute to disordered thresholds, lending further support to the idea that this may be a significant measurement issue affecting the internal validity of scales.⁵⁴

The dependency issue, revealed by the high residual correlations between items, is a further indication that more attention is required in relation to scale content. The problem with item dependency is that it inflates artificially the reliability of rating scales (as measured by PSI value), and therefore in measurement terms, makes a scale look better than it actually is. One could argue that some degree of item dependency might be expected for these 9 rating scales, because many needed to include tasks assessed bilaterally so that clinically significant variations could be addressed therapeutically (e.g., development or worsening of contractures). For example, the HFMS includes items that relate to rolling prone to supine on the left and right sides (items 6 and 7, respectively), and the NSAA includes items that ask about standing on the right and left legs (items 3 and 4, respectively), activities where dependency might be expected. These important issues need to be taken into consideration in the event that subsequent attempts to improve scale content are undertaken. This does not mean that items must be deleted from a scale in the future, but they may be excluded from a transformed measure or considered as a sub test.

DIF is a key issue in Rasch analysis.⁵⁵ The Rasch model is based on the requirement of invariance (stability).³⁷ This does not mean that examining data with the Rasch model automatically produces stable results. It means that, with the Rasch model, we are able to test if the items of our scales are stable across the trait they measure and the clinically different groups in which they are used. All of the scales assessed in this study exhibited DIF to a certain degree across SMA categories. This suggests there is a degree of instability of scale scores generated in the different groups of patients. The clinical implication is that measurement is affected to an unknown degree, and the direct comparability of scores between groups is uncertain. Once again these are pertinent issues that need to be taken into consideration when considering potential scale improvements. We would stress that, while ensuring scale stability is important, other issues relating to scale sufficiency, such as correctly ordered thresholds, should be corrected first.”

Supplementary Table S2 also reports findings related to scale stability over type of SMA, age, and group. Although ensuring scale stability is important, other issues related to scale sufficiency, such as correctly ordered thresholds, must be corrected first.

The final key findings related to targeting issues were identified for some of the scales. The EK2 (nonambulant), HFMS (nonambulant), and GMFM (nonambulant and ambulant) were found to be better targeted to less affected children within the context of their target population, and the MFM (nonambulant and ambulant) and NSAA (ambulant) were better targeted for more affected children within the context of their targeted populations. This suggests that the match between the range of motor performance measured by these scales and the range of motor performance measured in these samples is suboptimal. This increases the error around the measurements of individuals, reduces the ability of the scales to discriminate accurately between different levels of motor performance, and limits the ability to detect change over time as the disease progresses. This finding further indicates that additional consideration is needed for scale content and its defined purpose within a clinical trial. One potential consideration is to increase the measurement range of these scales by including additional items that better capture the range of problems for each SMA type. This is especially important when considering appropriate measurement precision across the whole of scales to adequately capture clinical change when it occurs.

The results lead to 3 main recommendations related to the potential next steps for further psychometric research on SMA scales. First, the content of each scale should be reviewed and modified in terms of misfitting items and item response categories with reversed thresholds. Second, most of the scales would benefit from an expert review of items that appear dependent on each other. Third, the content of the EK2, GMFM, HFMS, MFM, and NSAA should be reviewed, and additional items should be considered for people in the more/less affected range of SMA. A reasonable first step in further analyses of prospective datasets is a post hoc analysis of existing data, in which item response options can be changed and items removed or added. This is possible with the RUMM2030 software used in this study, and just such an analysis is planned. Findings from this analysis however should only be used to guide the next step, because any proposed alterations to scales would necessitate the need for further psychometric validation studies to assess the impact of the changes, either through natural history studies or baseline trial data.

Another important step that would strengthen the outcome scale would be to review explicit construct definitions of exactly what is being measured by each scale. This additional substantive information would make scale data more interpretable clinically.⁵³ In fact, a clear understanding of the substantive theory underpinning rating scales is central to valid measurement,^53,56,57 a fact now recognized and stressed by regulatory bodies such as the FDA.^7,9,38,58 We are not suggesting that any of the scales in this study are clinically meaningless. Quite the contrary; there is little doubt that each has been constructed carefully to provide valuable information that can be used for clinical assessment. However, our focus here is on measurement as opposed to other more evaluative uses of rating scales. These have different requirements. Thus, measurement has a very specific meaning with respect to quantifying clinical attributes (i.e., a quality, characteristic, or property belonging to a person). In contrast, the use of scales as clinical assessment tools is often a qualitative, information sharing process. When a rating scale is used as an outcome measure, it is being used as a measuring instrument. This has implications both clinically and psychometrically. We discuss these issues more fully elsewhere.¹¹

Construct definitions are elaborated as construct theories which, simply stated, are “the story we tell about what it means to move up and down the scale for a variable of interest” (e.g., temperature, reading ability, short-term memory). Why is it, for example, that items are ordered as they are on the item map? Does the ordering match clinical expert opinion?⁵⁹ The development of such theories should include evaluations of the extent to which the items of each of the scales mark out the construct (i.e., motor performance) to be measured, establishing the most appropriate item wording, structuring the scoring accurately, and conducting qualitative cognitive debriefing interviews to ensure consistency in meaning. Thus, how each of the scales capture motor performance in SMA in terms of ambulatory performance, standing, and transfers, and nonambulatory performance including sitting and lying or upper limb and distal performance, needs to be clarified explicitly. An exciting potential outcome from this process is that it provides a platform to compare the conceptual frameworks of each of the 9 rating scales, a process that in turn could lead to a clinical item bank⁶⁰ which has the potential to unify outcome measures across SMA clinical trials. Such an item banking approach, combining Rasch measurement theory with detailed construct theory, has proven to be extremely successful in other areas of human sciences.^59,61

This study has 2 main limitations. First, the datasets provided for the CHOP-INTEND, EK2, MFM, NSAA, and TIMPSI were smaller than the widely recommended n > 300. Therefore, to confirm their psychometric properties, these scales require further evaluations in larger datasets. Second, validity testing for all scales was limited in the extent to which construct validity could be assessed, especially in the comparison between different scales. Further examinations would be beneficial and would require additional data points (assessments).

These findings have important implications for the results of SMA clinical trials and research, and have fundamental implications for the development of better outcome measures from rating scales currently used in SMA. Rasch measurement methods add sophistication to traditional psychometric methods, providing detailed diagnostic item-level data. By highlighting and elaborating the key problems of the items of the 9 motor performance rating scales, these analyses provide an initial evidence base for future research to improve these scales. In addition, building detailed construct theories and allowing for comparisons across each of the rating scales provides a unique opportunity to bring together the scales in the form of an item bank that will add precision, reduce patient and clinician burden, and give clinical researchers a common metric to compare and contrast the outcomes of new therapeutic trials. Clinical trials under development using current scales can benefit from this overarching technique by relating raw scores collected on current scales to the metric of measures developed in the future by combining items and scales.

Abbreviations

AmSMART

American SMA Randomized Trial group

CHOP-INTEND

Children’s Hospital of Philadelphia test of neuromuscular disorders

DIF

differential item functioning

EK2

Egen Klassifikation Scale v.2

FDA

Food and Drug Administration

FitRes

fit residual

GMFM

gross motor function measure

HMFS

Hammersmith Functional Motor Scale

HFMSE

Expanded Hammersmith Functional Motor scale

ICC

International Coordinating Committee for SMA Clinical Trials

MHFMS

Extend, Modified Hammersmith Functional Motor Scale Extend

MFM

Motor Function Measure

NSAA-SMA

North Star Ambulatory Assessment for SMA

PNCRN

Pediatric Neuromuscular Clinical Research Network for SMA

PSI

Person Separation Index

RMM

Rasch Measurement Model

SMA

spinal muscular atrophy

SMAF

SMA Foundation

SMN

Survival Motor Neuron

TIMPSI

Test of Infant Motor Performance Screening Items

TREAT-NMD

Translational Research in Europe – Assessment and Treatment of Neuromuscular Diseases