Psychometric Characteristics of the Motor Function Measure in Neuromuscular Diseases: A Systematic Review

Abstract

Background:

Recent pharmaceutical breakthroughs in neuromuscular diseases may considerably change the prognosis and natural history these diseases. The ability to measure clinically relevant outcomes such as motor function is critical for the assessment of therapeutics and the follow up of individuals. The Motor Function Measure (MFM) is a quantitative scale designed to measure motor function in adult and children with neuromuscular disease (NMD).

Objective:

The objective of this study is to assess the quality and level of evidence of the MFM’s published measurement properties by completing a systematic review of the validation and responsiveness studies of the MFM20 (a 20-item version of MFM adapted for children 2 to 6 years of age) and the MFM32 (the original 32 item version), in all NMDs and in specific diseases.

Methods:

A search for MFM responsiveness and MFM validation studies was completed in February 2023 in EMBASE, MEDLINE, SCOPUS and Web of Science databases. The PRISMA guidelines and the COSMIN manual for systematic reviews were followed for databases searches, articles screening and selection, study quality and measurement properties evaluation.

Results:

49 studies were included in analysis. In studies including individuals with all NMDs, MFM’s internal consistency, reliability, convergent validity, construct validity and responsiveness were rated as sufficient with a high quality of evidence. Structural validity was rated sufficient with a moderate quality of evidence In SMA in particular, MFM’s reliability, internal consistency, convergent validity, discriminant validity and responsiveness are sufficient with a high quality of evidence. More studies would be required to assess specific measurement properties in different diseases. MFM32’s minimal clinically relevant difference has been defined between 2 and 6%.

Conclusion:

MFM’s structural validity, internal consistency, reliability, construct validity, convergent validity and responsiveness have been verified with moderate to high level of evidence.

INTRODUCTION

The last decade has brought many pharmaceutical breakthroughs, especially in the area of neuromuscular diseases (NMD). These new treatments may considerably change the prognosis and the natural history of diseases, an example being spinal muscular atrophy (SMA) with the use of nusinersen, onasemnogene abeparvovec, or risdiplam [1]. Since motor function represents a clinically relevant outcome for individuals and families and is the direct target of new treatments in NMDs, the ability to accurately measure this outcome is critical. Disease-specific and/or ambulatory ability dependent tools, for example the Hammersmith Functional Motor Scale for non-ambulant and ambulant individuals with SMA [2, 3], or the North Star Ambulatory Assessment, for ambulant boys with Duchenne muscular dystrophy (DMD) [4] have been developed. Other scales which focus on specific components of motor function include the Vignos [5] and Brooke [6] scales. The use of multiple tools in daily clinical and research practice can be time- and resource-consuming, require specific training, and most importantly can be fatiguing for the participant. The use of a single tool for motor function, validated in all neuromuscular diseases, could be more sustainable in research and clinical practice, without altering the efficiency of the assessment.

The Motor Function Measure (MFM) is a quantitative scale designed to describe motor abilities in individuals with NMDs). The MFM32 is composed of 32 items scored using a 4-point Likert scale based on the individual’s maximal abilities without assistance. Total score is obtained by summing the items and expressed as a percentage to allow comparison with other scores. The proof of concept MFM32 study was published in 2005 by Bérard et al., including 303 persons with NMD [7]. The 32 item MFM was validated in individuals 6 to 60 years of age. A 20 item adaptation of the MFM was validated in 2013 [8], for children from 2 to 6 years of age. The MFM consists of 3 domains: standing position and transfers, axial and proximal motor function and distal motor function. The MFM20 and MFM32 can be used regardless of severity of the disease or ambulant status. In clinical practice, the MFM is used to monitor disease evolution and allow caregivers to communicate with each other while defining rehabilitation interventions.

Psychometric properties of a scale define the ability of the instrument to measure what it aims to measure, and to reproduce a result consistently in time and space. In designing clinical trials, it is critical to choose the correct tool. The quantification of clinical parameters like motor function requires the certainty that the tool used has strong psychometric properties, given the possible negative consequences if there are errors in daily clinical assessment of a person or treatment efficacy in research. As defined by the Consensus-based standards for the selection of health measurement instruments (COSMIN), these properties include: structural validity(dimensionality of the construct), internal consistency (the interrelatedness among the items), cross-cultural validity, reliability, measurement error, convergent validity, construct validity and responsiveness [9, 10]. Since MFM32’s initial publication in 2005, many validation and natural history studies in adult and pediatric diverse populations have been published. As recent as 2022 Andrade et al. [11] recommended the use of the MFM as a valid, standardized tool for the assessment of activity in muscular dystrophy. While the MFM was initially validated in a diverse NMD population, recent publications took interest in its validation properties in specific populations, such as SMA and DMD.

The aim of this study is to assess the quality and the level of evidence of MFM’s published measurement properties, in all NMDs and in specific diseases.

MATERIALS AND METHODS

The study protocol was registered on the International Prospective Register of Systematic Reviews (PROSPERO) (registration number CRD42021261491). The Preferred Reporting Items for Systematic Reviews and Meta-analyses 2020 (PRISMA 2020) guidelines, and the COSMIN methodology for systematic reviews for outcome measures and patient reported outcome measures (PROM) were followed [9]. Data sources searches, title and abstract screening, assessment of the remaining full text, studies selection, data extraction, methodological quality assessment of each study and the grade of evidence were completed independently by two investigators (PR and SR). Discrepancies were resolved by discussion between PR, SR and a third investigator, CV.

Data search, eligibility criteria and study selection

EMBASE, MEDLINE, SCOPUS and Web of Science databases were searched in February 2023. Inclusion criteria were all validation and responsiveness studies in NMDs published from January 1, 2005 onwards. Validation studies are defined as studies reporting results about MFM20 or MFM32’s structural validity, internal consistency, cross cultural validity, reliability, measurement error, convergent validity, construct validity. Responsiveness studies are all studies reporting results on MFM’s sensitivity to change, including natural history studies using MFM as an outcome measure. Publications without full text English, reviews, study protocols, and/or recommendations were excluded. Studies using MFM as an outcome (interventional and cross-sectional studies) but with no results on MFM’s measurement properties were not included. In interventional studies reporting results on MFM’s responsiveness, only the data regarding the placebo group were interpreted. Regarding responsiveness, studies on DMD patients which did not report results on steroid naïve patients were excluded. Indeed, the long-term modification of the course of the disease in treated patients [12] would alter the interpretability of the results regarding MFM responsiveness.

The search string included a combination of keywords regarding neuromuscular diseases and the Motor Function Measure from 2005 onwards. The search strings for each database are detailed in supplementary data (supplementary material 1).

Duplicates and additional records identified through other sources (cross referenced) were eliminated. A first selection was performed by the screening of titles and abstracts. The full-texts of studies retained after this screening were assessed. Full-texts responding to the pre-specified inclusion criteria were included in qualitative analysis.

Data extraction

The following data were extracted for all studies: publication (title, authors, journal, and pages, year), study registration number (e.g., clinical trial number), type of study, participants characteristics (pathologies of individuals, number of ambulant, non-ambulant and total individuals, age of individuals range if available), and version of MFM used (32 or 20). In validation studies, data extracted included: structural validity (factorial analysis, Rasch analysis) internal consistency (e.g., use of Cronbach alpha), intra-rater reliability (e.g Intraclass correlation coefficient (ICC) or kappa), inter-rater reliability (e.g., ICC or kappa), convergent validity (e.g. Spearman or Pearson coefficients), construct validity, comparison of groups or not. In responsiveness studies, we looked for the primary results related to MFM scores in the evolution of disease, the minimal clinically important difference (MCID), and any group comparison. Data collection was performed using a systematic Excel form.

Methodological quality assessment

Risk of bias and assessment of the quality of the studies for each statistical property was assessed using the COSMIN Risk of Bias Checklist (July 2018) [10]. Criteria 1 and 2 were not applicable since MFM is not a patient-reported outcome, subsequently criteria 3, 4, 5, 6, 7, and 9 were used to rate each manuscript. Due to the absence of a well characterized gold standard for motor function measurement, criterion validity could not be assessed according to the COSMIN manual for systematic reviews of Patient-reported outcome measures (PROM). Criterion 10 was used to assess responsiveness but an adapted statistical method was used: studies calculating area under the curve (AUC) or using a correlation with another gold standard were classified as “Very good”. Studies using Standardize Response Mean (SRM) and/or the measure of effect size were classified as “Adequate”. SRM and effect size do not appear in the box 10 of the checklist, however it is mentioned in the COSMIN manual for systematic reviews that “expected effect size or similar measures such as SRM can also be used, [...] when an explicit hypothesis (and rationale) for the expected magnitude of the effect size is given” [10]. Depending on the value of SRM, responsiveness is assessed as large (SRM >0.8), moderate (SRM 0.5–0.8) or low (SRM 0.2–0.8). Studies performing other comparison methods or showing minor methodological flaws (assessed by the investigator) were classified as “Doubtful”. Studies showing important methodological flaws (assessed by the investigator) were classified as “Inadequate”.

Rating of measurement properties

Measurement properties of MFM were rated as sufficient (+), insufficient (−) or indeterminate (?) according to the COSMIN updated criteria for good measurement properties [10] for each study, and for every measurement property after synthesis of the results.

Rating of the quality of evidence

After rating the measurement properties of MFM, the grade of evidence was assessed for each measurement property using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach as detailed in the COSMIN guidelines [10].

RESULTS

Results of the screening and inclusion processes are reported in Fig. 1 (flow chart). Over the 178 full texts assessed for eligibility after title and abstract screening, 129 were excluded: 59 were cross sectional studies, 32 had no result regarding MFM’s psychometric properties, 4 had no results on MFM, 15 were not original studies, and 18 were responsiveness studies giving no information on untreated patients, or corticosteroids <di>näive DMD patients.

Fig. 1.

PRISMA 2020 flow diagram for new systematic reviews which included searches of datebases, registers and other sources.

Forty-nine studies were included in total analysis, including 16 validation studies and 37 responsiveness studies. 4 studies overlapped as both, validation and responsiveness studies (Fig. 1: flow chart).

The results are summarized in Table 1. A summary and analysis of published evidence by diagnosis is outlined in Table 2. Information about the included studies is detailed in supplementary material 2. Detailed results for each measurement property are displayed in supplementary material 3

Table 1.

MFM32 and MFM20’s psychometric properties and quality of evidence

Measurement property (number of studies)		Summary of findings	Rating	Quality of evidence
Structural validity (6)		Structural validity affirmed by two very good quality studies (MFM32 and MFM20) but without reporting of CFI, TLI or RMSEA or SRMR CFI >0.95 with CFA in DM1 and FSHD, = 0.92 in CMT in 1 very good quality study including 911 individuals about MFM32	Sufficient (+)	Moderate
Internal consistency (9)		Cronbach α range: MFM32 total 0.89–0.99, D1:0.87–0.98, D2:0.90–0.96, D3:0.69–0.89, acceptable to excellent. MFM20 total 0.97, D1 0.96, D2 0.90, D3 0.69, acceptable to excellent.	Sufficient (+)	High
Cross-cultural validity (3)		Studies for MFM32 validation in Portuguese, Turkish, Chinese. No CFA, IRT or regression analysis performed	Indeterminate	Low
Reliability (11)		ICC range inter-rater: MFM32: Total 0.92–0.99, subscores 0.78–0.99. MFM20: Total 0.99, subscores 0.91–0.99 Test-retest: MFM32: Total 0.93–0.99, subscores 0.785–0.99. MFM20: Total 0.99, subscores 0.91–0.99	Sufficient (+): ICC ≥0.70	High
Construct validity (17)	Convergent validity (11)	Significant correlations between MFM32 and MFM20 total and subscales and other validated motor function scales (Brooke, Vignos),and between MFM32 and VAS (visual analog scales for severity of disability), Clinician global impression in all NMD. Correlation with Brooke and Vignos ≥0.7 in at least two good quality studies including more than 400 individuals in total for MFM32, in 1 very good study including 88 individuals for MFM20.	Sufficient (+)	High
	Discriminant validity (6)	Strong inverse correlation between the severity of disability for MFM32 and MFM20, as assessed by physicians in 1 very good study (88 individuals) for MFM20, in 1 good quality study (303 individuals) for MFM32. Good discrimination for MFM20 D1, D2, and total score between the ambulant and nonambulant individuals (P < .001 for all scores) in one very good quality studies (88 individuals); for D1, D2, D3 and MFM32 total score in another study with adequate methodology including 33 individuals.	Sufficient (+) More than 75% of the results are in accordance with the hypotheses	High
Responsiveness (29)		In very good or adequate studies: Significant change in at least 3 study about SMA (126 individuals with MFM32, 81 individuals with MFM20 and MFM32), 2 study about all NMDs (171 individuals, MFM32), 1 study about dysferlinopathy (160 individuals, MFM20 in individuals older than 6 years old), 2 studies about DMD (123 individuals, MFM32), In adequate studies: no significant change in 1 study about LGMD2A (85 individuals, MFM32), 1 study about CMT (233 individuals, MFM20 and MFM32), 2 studies about SMA (33 individuals, MFM32). In doubtful studies: no significant change in 4 study about SMA (503 individuals, MFM32), 1 study about GSDIII (18 individuals, MFM32), 5 study about DMD (141 individuals, MFM32), 1 study about RYR 1 (47 individuals, MFM32), 1 study about post-polio syndrome (33 patients, MFM32), 1 study about oculopharyngeal muscular dystrophy (MFM32, 44 patients). In doubtful studies: significant change in 1 study about X linked myotubular myopathy (41 individuals MFM20 and MFM32), in 1 study about GSDIII (13 individuals, MFM32), 3 studies about SMA (243 individuals, MFM32), 2 studies about COL6 RD and LAMA2RD (91 individuals, MFM32) With anchor-based methods, for MFM32: MCID evaluated at 5.3% per year in all NMD (Vuillerot 2012), 4.7% to 6.1% in DMD (Huang 2021 – DMD treated and untreated individuals), 2.5% to 3.9% in LAMA2-RD and Col6-RD (Le Goff 2020). In SMA, LS mean change with CGI = –5.3 points (CGI) or –5.4 points (PGI) for minimally worse or worse group (Trundell 2020)	Sufficient (+) More than 75% of the results are in accordance with the hypotheses	High

CMT: Charcot Marie Tooth disease, Col6-RD: Col6-related disease, DM1: myotonic dystrophy type 1, DMD: Duchenne muscular dystrophy, FSHD: facio-scapulo-humeral dystrophy, GSD: glycogen storage disease, LAMA2-RD: LAMA2-related disease, LGMD: limb-grindle muscular dystrophy, NMD: neuromuscular diseases, SMA: spinal muscular atrophy, RYR1-RD: RYR1 related disease.

Table 2.

Analysis of published evidence on MFM’s measurement properties depending on the diagnosis

	All NMD	DMD	DM1	SMA	FSHD	CMT	CMD, CM
Structural validity	+ Moderate	? Moderate	+ High	? No evidence	+ High Ceiling effect especially for D2 and D3	+ High	+ Moderate
Internal consistency	+ High	? Moderate	? No evidence	+ High	? No evidence	? No evidence	+ High
Reliability	+ High	? Low	? No evidence	+ High	? No evidence	? No evidence	Inter-rater: + Intra-rater: ? Moderate
Cross-cultural validity	Low	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence
Criterion validity	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence
Convergent validity	+ High	? No evidence	? No evidence	+ High	+ Moderate	? No evidence	+ High
Discriminant validity	+ High	? No evidence	? No evidence	+ High	? No evidence	? No evidence	+ High Floor and ceiling effect
Responsiveness	+ High	+ High	? No evidence	+ High	Low	CMT1: − CMT2:+ High	+ Moderate
Measurement error	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence	? No evidence

GRADE level of evidence: green: high, orange: moderate, red: low, grey: no evidence. Rating of measurement properties: +: sufficient, ?: indeterminate, –: low. NMD: neuromuscular disease, CMD: congenital muscular dystrophy, CM: congenital myopathy, DM1: myotonic dystrophy type 1, SMA: spinal muscular atrophy, FSHD: facio-scapulo-humeral dystrophy, CMT: Charcot-Marie-Tooth disease, DMD: Duchenne muscular dystrophy.

Validation studies

Of the sixteen validation studies, including individuals from 2 to 77.2 years old, eight included all NMDs. The remaining 8 studies included one with individuals with congenital muscular dystrophies (CMD) and congenital myopathies (CM), one with CMD (LAMA-2 related diseases (LAMA2-RD) and Col6 related diseases (Col6-RD)), one with Charcot-Marie-Tooth disease (CMT), myotonic dystrophy type 1 (DM1) and Facio-scapulo-humeral dystrophy (FSHD), one with only FSHD individuals, 2 with non-ambulant SMA types II and III individuals, one with ambulant SMA III individuals, and one with ambulant DMD patients. Six studies were multicentric, with nine being a single center study. Thirteen had prospectively recruited individuals, 2 were retrospective in nature,and 1 enrolled both prospective and retrospective individuals.One study used the MFM20 for the younger individuals, while two studies used MFM32 for individuals under 6 years old [13, 14]. One study specifically studied the use of the MFM32 in children under 6 years old [14] while the second study indicated that there was no difficulty in the use of MFM32 in children under 6 years old and used a home-based MFM method.

Structural validity

Structural validity was evaluated in 6 studies, of which three were rated very good. One adequate methodological study about MFM32, with Exploratory Factor Analysis in all NMDs [7], did not report the comparative fit index (CFI), the Tucker Lewis index (TLI), the root mean square error (RMSEA), or the Weighted Root Mean Square Residual (WRMR). One very good methodological study about MFM32 showed a confirmatory factor analysis (CFA) [15]. One very good methodological study confirmed structural validity of MFM20 in all NMDs using Principal Component Analysis (PCA) [8].

One very good methodological study and 2 adequate methodological study about MFM32 performing Rash analyses showed limitations in CMD and CM individuals, and in FSHD and ambulant DMD individuals with a ceiling effect, especially for D2 and D3 [16, 17].

The overall rating of structural validity was classified as sufficient (+), with a moderate quality of evidence (downgraded by the indirectness of evidence, since the CFI, TLI and RMSEA are only available in DM1, FSHD and CMT).

Internal consistency

Internal consistency was evaluated in 9 studies (7 very good methodological studies, 1 adequate methodological study, 1 doubtful methodological study). 6 very good methodological studies calculated Cronbach’s alpha for MFM total score: 1 study for MFM20’s, 4 studies for MFM32. [7,8,14,18–20]. Five studies (4 very good methodological studies and 1 doubtful methodological study) calculated Cronbach’s alpha for MFM’s subscales [7, 8, 19–21].The internal consistency was rated as high (Cronbach alpha being ≥0.70 for MFM32 and MFM20 total scales and subscale in several very good studies with consistent results).

Cross-cultural validity

Cross cultural validity was assessed in 3 inadequate methodology studies about MFM32: a Turkish version, a Chinese version, and a Portuguese version of MFM [20–22]. However, no CFA, item response theory (IRT) or regression analysis was reported for these 3 versions: cross-cultural validity is rated as indeterminate, with a low quality of evidence.

Reliability

Intra-rater reliability was evaluated in 2 very good methodological studies about MFM32 for MFM total score [14, 18], in 3 adequate methodological studies about MFM32 for total score and subscores [7, 20, 23], and in 1 doubtful methodological study about MFMF32 [21]. For MFMF20, Intra-rater reliability was evaluated in 1 adequate methodological study for total score and subscores [8]. Inter-rater reliability was assessed by 5 studies for MFM32 total score and subscores (4 adequate studies and 1 doubtful methodological study) [7, 13, 19–21]. For MFM20, 1 adequate methodological assessed inter-rater reliability for MFM total score and subscores [8].

Reliability is rated as sufficient (+), with ICC ≥0.70 for inter-rater and intra-rater reliability in all adequate or very good methodological studies about MFM32 and MFM20, with high quality of evidence.

Measurement error

Standardized error of Measurement, limits of agreement, and smallest detectable change were not explicated in any of the included studies. MCID was calculated in 3 studies [18, 24, 25] with anchor methods. However, it was not possible to assess whether the smallest detectable change is inferior to MCID or not. As specified in the COSMIN guidelines, the quality of evidence regarding measurement error cannot be assessed.

Criterion Validity

Criterion validity would require a gold standard for motor function assessment. As specified by the COSMIN manual for systematic reviews of PROM, we considered there was no reasonable gold standard for motor function assessment, even though MFM is compared to widely used and well-known instruments such as Brooke and Vignos scale: no published data can confirm the properties of these scales as gold standards. Therefore, quality of evidence of criterion validity cannot be assessed and the results of comparison with other measurement instruments are interpreted within the convergent validity section.

Convergent validity

MFM was significantly correlated with Vignos and Brooke scales in the 5 very good methodological studies: 1 study about MFM32 and 1 study about MFM20 using Pearson’s coefficient found significant moderate to high correlation between MFM20 and MFM32 total score and subscales and Vignos and Brooke scales [8, 21]. 4 very good methodological studies (with respectively 303, 100 and 165 participant) about MFM32 used Spearman coefficient and found high correlation of MFM with Vignos scale, and Brooke scale [7, 14, 26, 27]. In 1 adequate methodological study about MFM32, Spearman coefficient showed low to high correlation of MFM total and subscales with Vignos scale and Brooke scale [20]. Significant correlations between MFM32 and MFM20 total scales and subscales and VAS (visual analog scales for severity of disability) [7, 8, 20], and between MFM32 and Clinician global impression [7, 14], forced vital capacity in SMA II and III [18], and with the the Paediatric Outcomes Data Collection Instrument upper extremity, mobility and transfer, but not with the happiness domain [13]. In SMA type 2 and non ambulant type 3, MFM showed significant correlation with Myo-Grip, MyoPinch and MoviPlate baseline score [28] In LAMA2-RD and Col6-RD, MFM32 significantly correlated with forced vital capacity, 10m walk, myometry,Egen Klassification Scale Version2,stand time, run10m time, Ascend4 steps, descend4steps, 6minwalk, Quality of Upper Extremity Skills Test, Activlim, North Star Ambulatory Assessment, Hammersmith functional motor scale, PedsQL^™3.0 Neuromuscular Module, but not significantly with 6minwalk, Patient-Reported Outcomes Measurement Information System, PedsQL^™3.0 Neuromuscular Module child self-report [19].

Convergent validity of MFM32 and MFM20 was assessed as sufficient (+) with correlation with other outcome measurement (Brooke, Vignos) ≥0.7 in at least three very good methodological quality studies including more than 500 individuals in total, with a high quality of evidence.

Discriminant validity

Discriminant validity was assessed in 2 adequate and 4 very good methodological studies. 1 adequate methodological study about MFM20 and 1 very good methodological study about MFM32 found a significant correlation between MFM and clinical global impression (CGI) [7, 8]. MFM32 total score and subscores allowed a good discrimination between diagnosis groups (DMD, Becker muscular dystrophy (BMD), FSHD, CM, CMD, myotonic dystrophy, SMA, hereditary neuropathies (HN), p < 0.0001) [7]. MFM20 allowed a good discrimination between ambulant and non-ambulant individuals (p < 0.001) [8].

In 2 very good methodological studies about MFM32 (NMD and SMA types II and III individuals), Least square means followed expected patterns of Hammersmith functional motor scale scores, Vignos scores, ability to sit [14, 18].

Discriminant validity was assessed as sufficient with more than 75% of the results in accordance with the hypotheses, with a high quality of evidence.

Natural history studies: Responsiveness and Minimal Clinically Important Difference

Thirty-seven studies were included in this category, including patients from 0.3 to 86 years old. Four studies adequately used MFM20 under 6 years old and MFM32 after 6 years old [29–32]. One study inappropriately used MFM20 for individuals older than 12 years old to reduce participant fatigue and duplication [33]. One study used MFM20 for patient solder than 6 years old with no justification. One study used MFM32 for children between 4 and 6 years old with no justification [34].

Two studies concerned all NMD. ten studies included DMD individuals, 2 study BMD individuals, 3 studies Limb-grindle muscular dystrophy (LGMD) individuals, 1 studies FSHD individuals, 12 studies SMA individuals, 2 study congenital muscular dystrophy individuals (LAMA2-RD, Col6-RD), 1 study Charcot Marie Tooth disease individuals, 2 studies Glycogen storage disease type III (GSDIII) individuals, 1 study dysferlinopathy individuals, 1 X-linked myotubular myopathy (XLMTM) individuals, and 1 RYR1-related disease (RYR1-RD) individuals, 1 studypost-polio syndrome individuals, 1 study nemaline myopathies individuals.

Eleven adequate methodological studies calculated SRM for MFM: In all NMD, for MFM32, SRM showed large responsiveness in 1 study (n=19) [26], low responsiveness in another study (n=152) [24]. For SMA individuals, 4 studies calculated SRM: SRM range showed low (n=19, all SMA and n=14, SMA type III and IV) to large (n=112, all SMA) responsiveness at 1 year for MFM32[23,35,36].One study using adequately MFM20 and MFM32, calculated SRM showing low responsiveness for total score at 1 year,and moderate responsiveness at 2 years[31]. For DMD, 2 adequate methodological study calculated SRM with a 1-year follow-up (n=49) showing a significant responsiveness, for MFM32 total score [37]. Schreiber et al. showed a large responsiveness for MFM32 total score at 2 years (n=74) [12]. For dysferlinopathy, mean change of MFM20 in individuals older than 6 years old at 1 year (median (range)) was evaluated at –1 (–13.0–6.0, p<0.001) with showing low responsiveness for MFM total score (n=160) [33]. One study about CMT using adequately MFM32 and MFM20 found no significant change at 1 except for CMT2 individuals (large responsiveness) [29]. One study about LGMD2A found no significant change in MFM total score after 24 months for MFM32 [38].

An adequate methodological study took interest in the least square mean change in SMA, which was–5.3 points (Clinician Global Impression) or –5.4 points (Participant Global Impression) for minimally worse or worse group (MFM32) [18].

A doubtful methodological study assessed the MCID using different methods for MFM32 total score in DMD, in a mix of treated and untreated patients: 6% using the average change method, 5% using the change difference method, and 5% using the ROC curve method [25]. In LAMA2 and Col6 related diseases, Le Goff’s et al. assessed MCID for MFM32 at 2.5 using the within participant method, and 3.9 using a between participant method [39]. In all NMD, MCID was calculated at 5% per year [24].

Responsiveness of MFM was rated as sufficient (+), more than 75% of the results being in accordance with the hypotheses, with a high quality of evidence.

Other comments on MFM properties

MFM20 Scale completion duration was 26±8.5 minutes (range, 12–50min) in de Lattre et al. study [8], 36 min (range 8–75 min) for MFM32 [7].

A floor and ceiling effect were reported in Col6-RD and LAMA2-RD [19]. In FSHD individuals, Mul et al. and Guillot et al. reported a ceiling effect, especially for D2 and D3 [15, 17].

DISCUSSION

Since the development of the scale in 2005, there have been many published manuscripts related to the MFM. Given the increased use of MFM scale in clinical trials, it is timely to complete a systematic review to verify MFM’s strong psychometric characteristics. In populations including all NMD diagnosis, MFM32 and MFM20’s internal consistency, intra-rater and inter-rater reliability, convergent validity, discriminant validity have been confirmed as sufficient with a high quality of evidence through several validation studies. Structural validity was rated as sufficient, but the level of evidence was downgraded to moderate due to indirectness. The analysis by diagnosis showed sufficient reliability, internal consistency, convergent validity, discriminant validity and responsiveness with a high quality of evidence in SMA, and with a moderate to high quality of evidence in CMD and CM. For SMA, the use of MFM can be recommended the use of MFM for the assessment of motor function. For CMD and CM, MFM can be used with caution regarding a possible floor and ceiling effect. In other diagnosis, specific studies would be required to evaluate the measurement properties of MFM precisely in every neuromuscular disease. In studies including all NMDs regardless of the specific diagnosis, or to compare different diagnoses, the use of MFM can also be recommended given the published evidence of its measurement properties in all NMD populations. Regarding DMD, although there are no specific study about MFM’s properties for this diagnosis (except regarding responsiveness), we could expect the measurement properties of MFM to be similar to the all-NMD population. DMD is indeed the most prevalent disease in all-NMD population (27% of individuals in Bérard et al. study) [7].´ However, this would need to be confirmed in specific studies.

The absence of a true gold standard for motor function did not allow us to assess criterion validity. In accordance with the COSMIN guidelines, data related to correlation of MFM with other motor function assessment tools were taken into account as convergent validity results. However, criterion validity is an important element for the choice of an outcome measure, and 7 very good studies chose Brooke and Vignos scales as reasonable gold standard to assess criterion validity. In these studies, significant moderate to high correlation of MFM20 and MFM32 with the Brooke and Vignos scales (Spearman coefficient range 0.56–0.91, Pearson coefficient range 0.49–0.91 for total score and subscrores) [7, 8, 13, 14, 18, 19, 26].

MFM32’sresponsivenessandsensitivitytochange are also validated in numerous pathologies, and a MCID has been defined between 2 and 6% per year depending on the diagnoses included in the studies: the change in MFM scores depends on the rapidity of progression of the disease. Slowly progressive disorders do not induce a change of MFM score in a period of 1 year for SMA type III, SMA type IV or CMT1, or 2 years for LGMD2A. There was also no change in a post-poliomyelitis syndrome population [40], which is expected considering the stability of this disease. Longer longitudinal studies may be required to observe a change in these diseases. It is interesting to notice a significant change in MFM scores at year one in individuals with CMT2 individuals, validating that CMT2 is often more rapidly progressive than CMT1 [29]. Similarly, MCID for MFM32 was calculated at 5% in all NMDs [24], whereas Huang et al. calculated the MCID up to 6% in individuals with DMD [25]. However, this last study used the Chinese version of MFM. The cultural validity of this version of the scale being indeterminate, the results of this study on responsiveness should be considered with the caution that they might not be generalized to the English version of MFM. The range of MCID in this study could be explained by the rapid progression of the disease in individuals with DMD: it would be important to assess motor function more frequently to accurately capture changes and avoid an overestimation of the MCID. MCID value is an important parameter to take in account when considering the use of an outcome measures in interventional study. However, Huang et al. [25] and Molino et al. [41] demonstrated that the MCID is not only influenced by calculation methods, anchor choice, and the chosen threshold for success, but also by extrinsic factors. Sample size, population characteristics, disease entity and severity can modify the MCID. Molino et al. suggests that each interventional study should calculate the MCID adapted to the population of study [41]. The use of MCID intervals instead of single values may also be more relevant to take in account its variations.

Given the high inter-and intra-rater reliability of MFM32, we expect the MCID to be superior to the standard error of measurement (SEM). However, studies reporting SEM or Limits of Agreement would be required to ensure an acceptable measurement error for MFM.

Cross-cultural validation of MFM32 in Portuguese, Chinese and Turkish was evaluated. Unfortunately, no confirmatory factor analysis or regression analysis were performed for any of these MFM translations. Thus, the rating of these MFM translations for cross-cultural validity remains indeterminate, with a low quality of evidence. Further studies with CFA or regression analyses would be required.

The use of MFM was consistent with the administration methods developed by MFM study team. Misuses were mainly due to the administration to the incorrect age group, use of MFM32 in younger individuals [14, 33] or the use of MFM20 in individuals older individuals [34]. We insist on the necessity to perform a total MFM including its 3 domains so its measurement properties are verified. Regarding each domain of MFM, a lower responsiveness is noted for D3, as well as a floor and ceiling effect for Col6-RD and LAMA2-RD.

The duration of MFM assessment can be considered as acceptable in the included studies (26 min for MFM20, 36 min for MFM32), considering the evaluation of all domains of motor function allowed by MFM 3 domains. In daily practice, this evaluation is well tolerated by individuals. However, new interactive methods for MFM’s assessment are being developed such as MFM play: a digital playful module of MFM (NCT04435093), to increase the participation of individuals. Moreover, the use of a single generalist scale for different neuromuscular disease to assess motor function could be more sustainable for clinical departments, rather than the use of several disease-specific scales. Indeed, the use of each instrument requires a training and a regular practice to beneficiate from the tool’s measurement properties. Measurement properties of MFM could also allow comparison between different pathologies for the assessment of a non-pharmacological intervention for example.

One study described use of MFM administration after viewing the videos provided by MFM team [22], which may be a source of bias and measurement error.

Limitations

An important limitation of our study is the possibility of a selection bias in favor of MFM studies with favorable outcomes. Those with unfavorable conclusions may not have been specified in the title or abstract, bypassing our research algorithms.

Even if MFM appears to have sufficient metrological properties in most diseases, ages and phenotypes, we show some limitations especially with a floor and ceiling effect in FSHD, in CMD, in CM. For CMT, the responsiveness of MFM depends on the severity of the phenotype (insufficient for CMT1, sufficient for CMT2). In the most and the less severe phenotypes, MFM’s sensitivity to change can be altered. Across the life of an individual, its performances can also vary depending on the age, and the evolution of the disease.

In this systematic review, studies including treated patients, providing data on responsiveness were excluded. This mainly concerned studies including corticosteroids treated DMD, as corticosteroids are part of the standard of care of DMD for many years. In the specific context of DMD, corticosteroids modify the evolution of motor function with a nonlinear effect, even for individuals whose treatment has been stopped, and the mix of treated, past-treated, and treatment naïve patients could alter the interpretability of responsiveness results [12]. In patients treated with corticoids, many publications providing data on MFM are now available [27]. In SMA where the new standard of care include disease modifying treatments (nusinersen, onasemnogene abeparvovec, or risdiplam), no new data from untreated patients will be available.

Another potential limitation is the generalization of the use of MFM. Although the scale was validated in a wide diversity of neuromuscular diseases with various phenotypes, progression patterns and age ranges, the studies reviewed were mainly assessing the MFM in genetic diseases beginning in childhood. Considering the frequency and diversity of adult-onset neuromuscular disorders (amyotrophic lateral sclerosis, diabetic neuropathy, alcohol-related neuropathy, chronic inflammatory demyelinating polyneuropathy, myasthenia gravis, Lambert Eaton syndrome, inflammatory myopathies), the MFM’s quantitative statistical properties would need to be established in these additional populations. Ideally, all measurement properties would be substantiated in the specific disease populations.

CONCLUSION

MFM is a standardized assessment tool used globally. Its quantitative measurement properties (structural validity, internal consistency, reliability, construct validity, convergent validity and responsiveness) have been established in a diverse neuromuscular disease population. Especially, specific studies validated these properties in SMA with a high quality of evidence, and in CMD and CM with a moderate to high quality of evidence. Limitations of the tool include: lack of cross-cultural validity, measurement error not assessed in any of the validation studies, and criterion validity not established given the absence of a gold standard for motor function.

Currently, the use of the MFM for the assessment of motor function can be recommended in individuals with SMA and, in CMD and CM with caution related to a floor and ceiling effect. Additional studies are needed to establish the MFM’s quantitative measurement properties in other neuromuscular diseases, and to compare the properties of MFM and other outcome measures in NMDs.

The Motor Function Measure assessment scale can be used across a person’s lifespan, regardless of disease or ambulatory status. Given the number of new pharmaceutical developments in the field of NMD, it is imperative to evaluate and track clinically relevant motor outcomes to determine efficacy of state-of-the-art therapeutics using robust scales such as Motor Function Measure.

ABBREVIATIONS

AUC

Area under curve

BMD

Becker muscular dystrophy

CFA

confirmatory factor analysis

CFI

comparative fit index

CGI

Clinician global impression

CM

congenital myopathies

CMD

congenital muscular dystrophy

CMT

Charcot Marie Tooth disease

Col6-RD

Collagen 6 related-disease

DM1

type 1 myotonic dystrophy

DMD

Duchenne muscular dystrophy

FSHD

Facio-scapulo humeral dystrophy

GSDIII

Glycogen storage disease type III

HN

Hereditary neuropathy

ICC

intraclass correlation coefficient

IRT

item response theory

LAMA2-RD

LAMA2-related disease

LGMD

Limb-grindle muscular dystrophy

MCID

Minimal clinically important difference

MFM

Motor Function Measure

NMD

Neuromuscular diseases

PCA

Principal component analysis

PGI

Participant global impression

PROM

patient-reported outcome measures

RMSEA

Root Mean Square Error of Approximation

RYR1-RD

RYR1 related disease

SEM

Standard error of measurement

SMA

Spinal muscular atrophy

SRM

Standardized response mean

TLI

Tucker-Lewis index

VAS

Visual Analogic scale

WRMR

Weighted Root Mean Square Residual

XLMTM

X-linked myotubular myopathy