Duchenne muscular dystrophy (DMD) is a rare, progressive neuromuscular disease caused by mutations in the DMD gene, resulting in the absence of functional dystrophin protein.^{1, 2} Dystrophin is a large subsarcolemmal protein that has a key structural role in muscle fiber, protecting against damage during normal muscle contraction as part of the dystrophin-associated protein complex.^{3, 4} Absence of functional dystrophin protein results in progressive skeletal muscle weakness, cardiomyopathy, respiratory insufficiency, and, ultimately, premature death.^{1, 2}

The DMD gene is located on the X chromosome; therefore, it is inherited in an X-linked pattern.⁵ Because inheritance is recessive and males receive only one X chromosome, DMD mostly affects males.^5-7 The estimated prevalence of DMD in males is approximately 9000 to 12 000 in the USA, with a frequency of 7.1 per 100 000 males globally.^{8, 9} While evidence of muscle damage is present at birth in DMD,¹⁰ maturational development often masks progressive muscle damage in the early stages of life.⁹ The typical age at symptom onset is early childhood (<5 years) and includes delays in independent walking,¹¹ a characteristic pattern of rolling onto the stomach and using the hands to push up into a standing position to accommodate leg weakness,¹² toe walking, and enlarged calf muscles.¹³

Patient-reported outcomes (PROs) are measures important for determining the impact of disease from the patient's viewpoint, providing an understanding of the natural history of the disease, its progression, and the extent of the impact on a patient's daily life.^{14, 15} This is especially important for rare diseases such as DMD, given its progressive nature, diversity of manifestations, and impacts on individuals with the condition, making it challenging to capture the full spectrum of their experiences solely through objective measures.¹⁶ These measures frequently focus on health-related quality of life, which refers to the patient's overall perception of the physical, psychological, and social impact of illness and treatment on their life.^16-18 As symptoms of DMD begin to occur when patients are too young to self-report using a PRO measure, proxy reports can provide important information on the impact of the disease when the PRO measure is completed on behalf of the patient from the viewpoint of their family or caregiver(s).¹⁴ Research indicates that parents of children with disabilities tend to report lower health-related quality of life for their children than the children report themselves.¹⁶ In particular, agreement between parents and children was lower in social domains but reasonable in domains affecting physical activity and functioning.¹⁹ Nonetheless, the caregiver's perspectives on their child's quality of life are important for gaining a comprehensive understanding of the impact of the condition on the patient's well-being and daily life.¹⁶ In this context, several PROs have been used to assess the impact of DMD on health-related quality of life. These include generic preference-based measures, such as the EuroQol 5-domain and the Health Utility Index, both of which contain domains of relevance to DMD, such as mobility and self-care. However, both instruments are used primarily for health economic analyses. There are also two condition-specific measures—the Pediatric Quality of Life Inventory DMD Module and the recently developed DMD Quality of Life measure. It is important to note that there is limited evidence supporting the use of the Pediatric Quality of Life Inventory in patients with DMD.^{20, 21} Additionally, the DMD Quality of Life measure lacks established norms specific to the US population; as such, it has had limited use in clinical research studies to date.²²

The Patient-Reported Outcomes Measurement Information System (PROMIS) Health Organization measures are a set of generic person-centered instruments designed to evaluate the impact of a disease on several aspects of health-related quality of life, including physical, social, cognitive, and emotional function; they were developed for use in the general population, including children and adults.²³ The PROMIS Mobility (pediatric or PP) item bank focuses specifically on evaluating a range of physical mobility-related activities, such as walking, running, climbing stairs, and other lower-extremity skills, which are relevant to a child's daily life and functioning.²⁴ While these instruments were developed and validated using cutting-edge psychometric methods, it is still crucial to document the reliability, validity, and responsiveness of item banks in the specific context of use and the population of interest.^25-27

Although the generic PROMIS Parent Proxy (PP) Mobility item bank including 23 items was developed to measure caregiver perceptions of a child's mobility over the past week, the psychometric validity of measuring mobility in DMD has not been established. Therefore, the objective of this study was to evaluate the validity and psychometric performance of the generic PROMIS PP Mobility 23-item bank in a cohort of ambulatory children with DMD using the Rasch statistical analysis. This analysis is relevant given that the PROMIS PP Mobility item bank is currently being administered in DMD clinical trials.^{28, 29}

METHOD

RESULTS

Preliminary analysis

A total of 151 observations were collected for the Rasch analysis, split equally according to age group. The distribution of responses for each item can be found in Table S1. Neuromuscular physical therapists opted a priori to exclude four items from the subsequent Rasch modeling, which were considered not clinically relevant or not applicable to patients with DMD, resulting in 19 items for inclusion in the Rasch analysis. The four items excluded were: (1) ‘ability to stand on tiptoes’ because this is a compensatory mechanism in DMD pathophysiology, which increases with disease severity;³⁷ (2) ‘ability to turn their head all the way to the side’ because this action is typically not affected in ambulatory patients with DMD;³⁸ (3) ‘use of a walker, cane, or crutches’. These assistive devices are typically discouraged in patients with DMD because of the increased risk of falling associated with these mobility aids;^{39, 40} and (4) ‘ability to physically do the activities they enjoy the most’ because it was predicted to be insufficiently discriminative and a potential source of bias; owing to the progressive disabling nature of DMD and resiliency of children, patients tended to gravitate to more sedentary activities that can be enjoyed despite decreasing mobility.

The initial Rasch analysis showed that only 12 of the items had appropriately ordered and distinct thresholds (Figure 1). The significant item–trait interaction test (p < 0.001, χ² = 95.8 with 38 degrees of freedom) showed that the data did not fit the model and that the items were not working as expected across different levels of mobility in the patient sample.

Reorganization of response categories

Response categories were reorganized for 12 of the 19 items because they had disordered thresholds in the initial analysis or were disorganized after the initial round of changes, or because the sample sizes in some of the response categories were too small to be robust. Three items had their response options collapsed from the original five to four; nine items had their response options collapsed from the original five to three (Figure 2). The remaining seven items were unchanged to maximize the granularity of the analysis of the items deemed most clinically relevant for mobility and impact on daily life.

An example of probability curves for an initially disordered item is item 3 (‘keep up when playing with other kids’), before and after regrouping the thresholds, which is shown in Figure 3. In this case, response option 1 (‘with a lot of trouble’) was problematic and selected infrequently, resulting in item 3 having disordered thresholds. This was addressed by merging response option 1 with response option 2 (‘not able to do’), based on expert clinical opinion. Able to ‘do sports/exercise like other same-aged kids’ exemplifies the probability curves for an item for which the original five response categories yielded five ordered and distinct curves and is presented in Figure 4. In this example, five original response categories were already ordered so that each category contributed to the measurement.

Finalized Rasch results

Overall model fit

After regrouping the response categories, the Rasch model was rerun, yielding a non-significant p-value for the item–trait interaction (p = 0.143, χ² = 47.3 with 38 degrees of freedom), indicating good overall model fit of the data. The power of the items to discriminate between respondents with different mobility levels was assessed as excellent (Person Separation Index = 0.95). All items and persons showed good fit to the Rasch model expectations. The average item and respective person fit residual values (standard error [SE]) were − 0.55 (0.68) and − 0.29 (0.79), respectively, which is in line with the 0 (SE = 1) target values. Unidimensionality was satisfied.

Item locations

Item locations and fit values are presented in increasing order of difficulty in Table 1. As expected for a population of ambulatory pediatric patients with DMD, the easiest items include ‘move legs’, ‘walk across the room’, and ‘get out of bed without help’. The most challenging activities were ‘run a mile’ and ‘ride a bike’.

TABLE 1. Item locations and fit values.^a

Item number	Item description	Location	SE	Fit residual	p
I-04	Move legs	−2.96	0.19	0.13	0.471
I-10	Walk across the room	−2.71	0.23	−0.83	0.038
I-16	Get out of bed without help	−2.58	0.16	−0.54	0.466
I-09	Get into bed without help	−2.00	0.14	0.03	0.141
I-14	Get up from regular toilet	−1.53	0.13	−0.12	0.721
I-23	Use a wheelchair to get around	−1.45	0.19	0.1	0.742
I-11	Bend over to pick something up	−0.91	0.18	−1.33	0.083
I-18	Go up one step	−0.90	0.12	−1.02	0.374
I-05	Stand up without help	−0.44	0.18	−0.73	0.159
I-02	Get up from floor	0.23	0.12	−1.99	0.061
I-19	Carry books in backpack	0.23	0.16	0.48	0.126
I-12	Get in and out of car	0.30	0.17	−0.77	0.609
I-13	Walk more than one block	0.52	0.13	−0.45	0.932
I-15	Get down on knees without holding on to something	1.04	0.16	−0.42	0.609
I-01	Do sports/exercise like other same-aged kids	1.59	0.12	−1.79	0.329
I-07	Walk up stairs without holding on to anything	1.70	0.11	−0.70	0.439
I-03	Keep up when playing with other kids	2.08	0.14	−0.73	0.769
I-17	Ride a bike	2.72	0.17	0.37	0.158
I-20	Run a mile	5.07	0.24	−0.09	0.367

• ^a Items are ordered according to the level of difficulty on the mobility continuum, with the easiest items first. Abbreviation: SE, standard error.

Threshold map

Figure 5 shows the threshold map of the customized PROMIS PP Mobility measure, with items ordered according to level of mobility difficulty. All items now have ordered thresholds. The map shows that pediatric patients with DMD experience increasing difficulty performing tasks as their overall mobility decreases and indicates that loss of ability is not observed in the same way across all items.

Person locations

The average person–item location was 1.04 (range: −4.39 to 6.24), indicating that patients performed on average slightly better on the mobility function continuum than expected by the model. There were no extreme values and therefore no floor or ceiling effect.

Person–item threshold distribution

Figure 6 shows the person–item threshold distribution of the revised measure and demonstrates person–item alignment where a person's ability (i.e. level of mobility) and item threshold (i.e. difficulty) are plotted together. Overall, item locations are evenly distributed along the continuum. However, there is a noticeable gap in the histogram for location values above 2.5, which suggests potentially suboptimal targeting of the PROMIS PP items for high levels of mobility. Additional items might be needed to optimize the conceptual breadth of the construct under investigation and in turn facilitate differentiation among high-mobility respondents with DMD.

Local dependency

A relatively large (0.499) residual correlation was observed between items 9 (‘get into bed without help’) and 16 (‘get out of bed without help’). Although correlated, the ability to perform these two tasks relies on the use of different combinations of muscles; additionally, performance can vary depending on disease stage. Therefore, the timing of loss of the ability to get in or out of bed without help may differ, making both items uniquely informative about disease progression. All other correlations observed were within the acceptable range of the model. The residual correlation matrix is shown in Figure S2.

Exploration of differential item functioning

One small indication of uniform DIF was observed between the patient's age group and one item (item 11, ability to ‘bend over to pick something up’). The p-value was just above the threshold for significance after Bonferroni correction (p = 0.000842, threshold = 0.000877). No other DIF, uniform or non-uniform, was identified on that grouping variable. Given that these age groups are only a proxy for early and late ambulatory patients, they do not directly measure ambulatory function. Furthermore, to be sensitive to the level of disease progression of the individual with DMD, the measure was not administered to the caregivers of extremely late ambulatory or non-ambulatory patients. Thus, no definitive conclusions were drawn from these data. Specifically, a model was not fitted by splitting item 11 because the basis for splitting that item would not be robust.

DISCUSSION

To the best of our knowledge, this is the first study to evaluate the validity and psychometric performance of the PROMIS PP Mobility item bank in patients with DMD using Rasch statistical analysis. The resulting 19 items from the Mobility scale achieved good overall model fit, with all items and persons showing good fit to Rasch model expectations. The easiest items were ‘move legs’, ‘walk across the room’, and ‘get out of bed without help’; the most difficult items were ‘run a mile’ and ‘ride a bike’. This is consistent with DMD pathophysiology, which is characterized by a well-documented progression pattern of deterioration of muscle strength and function over time. As observed, the abilities expected to be lost last were the easiest to perform. Some residual correlations were observed between items; however, these were considered reasonable owing to the nature of the items. Threshold map results were also as expected.

While striving to adhere to the specifications of the Rasch model, the goal was also to retain maximum granularity in responses whenever feasible. The PROMIS PP Mobility measure assesses activities (unlike the North Star Ambulatory Assessment, which is more skill-based), so it is likely to benefit from a broader scoring range.⁴¹ For this purpose, seven items considered the most clinically relevant for assessing mobility impairments in DMD (items with a high degree of overlap with the functional assessment, such as the North Star Ambulatory Assessment) were left unchanged.⁴² Therefore, our approach focused on preserving the PROMIS PP measure's capacity to detect and capture nuanced changes in mobility while maintaining alignment with Rasch model criteria.

The results show that the customized measure was effective in measuring lower-limb physical function in DMD across all levels of mobility, except for the highest levels. The potential lack of sensitivity at the high end of the mobility spectrum is not a major limitation affecting the relevance of the measure, as most patients with DMD have reduced function compared with age-matched controls. However, this finding suggests that further evaluation of items that target high levels of mobility is warranted.

The published literature on PROMIS PP measures in DMD is limited, which precludes a comparison of our results with directly relevant literature. Furthermore, Rasch analysis has seldomly been applied in the development or validation of measures frequently used in DMD.^{20, 43-45} The Pediatric Quality of Life Inventory is among the most frequently used disease agnostic instruments in DMD clinical research,²⁰ but different versions have failed to demonstrate robust validity and reliability.^44-46 In contrast, the DMD Functional Ability Self-Assessment Tool and the North Star Ambulatory Assessment have demonstrated overall satisfactory psychometric properties.^{43, 47} However, based on the literature published to date, the DMD Functional Ability Self-Assessment Tool instrument is underused in research and practice. For other generic instruments, such as KIDSCREEN and the 36-item Short-Form Health Survey, the quality of the published evidence on content validity in DMD has been assessed as low and very low, respectively.²⁰

This study has some limitations. We acknowledge a selection bias in that the authors felt that it was insensitive to give the PROMIS PP Mobility item bank to individuals who were no longer ambulant or could walk only a few steps, even though the caregiver could have answered the ‘my child can move his legs’ item. Although all patients had a confirmed genetic diagnosis consistent with DMD, a minority may have had a milder phenotype known as an intermediate or Becker-like phenotype.⁴⁸ However, given that this represented a fraction of patients, it is unlikely that it would have had a meaningful impact on our findings. Because of the rarity of DMD and the fact that all participants received specialty care at the same neuromuscular clinic, sociodemographic information was not included to prevent inadvertent identification of patients. We could not properly investigate DIF in our analysis; therefore, our model does not reveal whether some items may function differently among different factor groups that otherwise share the same ability estimate on the construct.

The present study has several strengths. The use of Rasch modeling allowed for a sophisticated psychometric analysis to be performed on the generic item bank to ensure the accurate measurement of mobility in ambulatory children with DMD. Having a measure fit for purpose with reliable scores would enable PRO measures to be more readily used in clinical research. By focusing on caregiver assessment, the item bank allowed gathering of insights into a young child's daily functioning, abilities, and challenges that could otherwise be difficult to collect, especially when a child is too young to self-report.¹⁴ The number of observations available for the analysis represented a relatively large sample for a rare disease like DMD. Results from the analysis are highly relevant given the lack of robust, content-valid, PROs in DMD clinical research. The item bank targets mobility, which is crucial for understanding the progression of the disease in ambulatory individuals and for evaluating the effectiveness of treatments, given that gradual loss of ambulation is a characteristic of DMD.¹

The findings of this study have important implications for clinical research and practice. The practical utility of a well-constructed measure is paramount to measurement success, especially in DMD clinical trials, where responses to treatment may be nuanced. The PROMIS PP Mobility measure may not only aid in the development of sensitive measures for capturing nuanced changes in clinical trials, but also in guiding clinical care and facilitating proactive planning for the evolving needs of patients with DMD. For instance, discussions around the eventual need for a wheelchair can be performed with a clear timeline in mind, allowing for proactive planning and potentially facilitating practical and emotional adjustments.

In addition to further validation of the 19-item measure in another independent sample of ambulatory patients with DMD, future research should evaluate the extent to which the PROMIS PP Mobility scale alone is suitable for use in patients with minimal symptoms, or whether the scale may benefit from supplementation with additional items to target the high end of the mobility spectrum in DMD.

CONFLICTS OF INTEREST STATEMENT

The authors have stated that they had no interests which might be perceived as posing a conflict or bias.