Abstract
Background and Objectives
Development of a prognostic score for time to loss of ambulation (LoA) may help guide treatment of patients with Duchenne muscular dystrophy (DMD). Information from clinical data sources was used to create a prognostic score for LoA for use in practice.
Methods
Data used included 4 natural history databases and 3 clinical trial placebo arms. Inclusion criteria were as follows: (1) 6-minute walk distance ≥75 m; (2) a subsequent visit with an outcome assessment that identified LoA; (3) current corticosteroid use; and (4) available data on prognostic factors. Classification and regression tree models for time to LoA were used to classify patients into 5 risk groups (group 5 = highest risk), prioritizing the easily obtained rise from floor (RFF) and 10-m walk/run (10MWR) times as predictors. The prognostic score was validated in the Cooperative International Neuromuscular Research Group data.
Results
Data from 608 boys were analyzed (mean age 9.1 years). Kaplan-Meier curves for time to LoA for each risk group were well separated, describing meaningful differences in time to LoA. Median time to LoA for risk group 1 was not reached during a median of 2.0 years of follow-up. Median times to LoA for risk groups 2–5 were 4.4 (95% CI 4.0–inf), 3.0 (2.1–4.1), 2.0 (1.6–2.1), and 0.9 (0.8–1.6) years, respectively. These risk groups showed concordant times to LoA among 226 patients in the validation data and provided finer separation of patients across different levels of LoA risk than thresholds based on individual functional tests.
Discussion
Patients with DMD were classified into 5 risk groups for LoA using a simple, easy-to-use prognostic score based on RFF and 10MWR thresholds. The prognostic score replicated well in validation data and performed better than previously proposed classifications used in practice. Incorporating the prognostic score for time to LoA into a clinical visit may allow clinicians to counsel patients on trial eligibility, need for mobility devices, home adjustments, financial accommodations, and psychosocial needs. The prognostic score may also be used in trials to inform inclusion criteria and stratification. Study limitations included limited data for prediction of LoA risk beyond 4 years.
Introduction
Duchenne muscular dystrophy (DMD) is a progressive neuromuscular disorder caused by sequence variations in the X-linked dystrophin gene, affecting 1 in 3,600–6,000 male births.1,2 Initial diagnosis is made at approximately 4–5 years of age, and affected individuals manifest declines in ambulatory function during childhood.1 Standard-of-care corticosteroids, including prednisone and deflazacort, and the novel corticosteroid vamorolone affect disease progression, but ultimately, muscle strength deteriorates and respiratory and cardiac complications emerge, contributing to mortality by early adulthood.1,3-5 Optimal standards of care positively affect disease progression,6 but there is no cure for DMD. Novel dystrophin restoration therapeutics, including exon skipping therapies, stop codon readthrough medications, and, more recently, adeno-associated viral gene therapy, have received full or conditional approvals in some countries, but these target a fraction of individuals with DMD. Clinical trials are ongoing to further investigate the safety and efficacy of these products, and of additional therapeutic strategies.7
Loss of ambulation (LoA) is a cardinal milestone in the progression of DMD.3,8 In treated patients, LoA typically occurs between the ages of 7 and 18 years, with a median age of 12–13.4 years.2,3,9 Variation is linked to the use, type, and dosing regimen of corticosteroid treatment, dystrophin genotype, and genetic modifiers.10 Age at LoA and time to LoA are reliable indicators of the severity of disease progression and predictors of other major disease milestones, such as loss of upper limb function and need for ventilatory support.3
Prognostic factors of changes in function in DMD have previously been identified. Patient characteristics combined with 10-m walk/run (10MWR) and rise from supine have been used to predict 1-year changes in measures of ambulatory function in DMD (6-minute walk distance [6MWD], timed 4-stair climb [4SC], North Star Ambulatory Assessment [NSAA] total score).11-13 Peak obtained NSAA scores and rise from supine between 6 and 8 years of age,14 as well as percentiles of NSAA and timed function tests,15 have been shown to be predictive of LoA. However, to our knowledge, prognostic indicators have not been incorporated into a score that can predict time to LoA. Development of a prognostic score for time to LoA may help guide treatment of patients with DMD, individualize care, provide anticipatory guidance for patients and families, and inform drug development by allowing stratification of patients for inclusion in clinical trials and contextualization of long-term outcomes.
This study used data from multiple clinical sources on ambulatory boys with DMD who had initiated corticosteroids but were not receiving other drug therapies for DMD to create a prognostic score for LoA for use in clinical practice. A data-driven, machine learning approach in conjunction with clinical expertise was used to derive the prognostic score, which was separately validated using external data. The prognostic score represents a standardized, validated, and simple method that clinicians can use to facilitate optimal patient care and for clinical trial purposes.
Methods
Data and Patient Population
Data used to develop the prognostic score were derived from 4 real-world and natural history data (RWD/NHD) sources and 3 clinical trial placebo arms (eTable 1). The data sources included those with patient-level data from ambulatory boys with DMD on the study measures required for development of a prognostic score for LoA, which were accessible through the collaborative Trajectory Analysis Project (cTAP) at the time of this study. RWD/NHD data included the Leuven Neuromuscular Reference Center, the PRO-DMD-01 prospective natural history study, the iMDEX natural history study, and the ImagingDMD natural history study.
Clinical trial placebo arms were obtained from Eli Lilly's phase 3 trial of tadalafil (NCT01865084) and from PTC Therapeutics' phase 2b and phase 3 trials of ataluren (NCT00592553 and NCT01826487). Data used to validate the prognostic score were RWD/NHD derived from the Cooperative International Neuromuscular Research Group (CINRG) Duchenne Natural History Study (DNHS).
Standard Protocol Approvals, Registrations, and Patient Consents
Data sources were approved by the ethics committees of each institution. All patient consent and approvals from institutional review boards were completed in the original studies. Only anonymous, deidentified data were analyzed in this study.
Sample Selection Criteria
Inclusion criteria for the development and validation samples were (1) 6MWD ≥75 m at a recorded clinic visit; (2) a subsequent visit with an outcome assessment that identified LoA (6MWD = 0 or “can't perform” in the development sample, wheelchair dependence in the validation sample); (3) current corticosteroid use (prednisone or deflazacort; both daily or intermittent dosing permitted); and (4) available data on candidate prognostic factors easily obtained in the clinic (age, timed function tests [10MWR, 4SC, rise from floor (RFF)], height, weight, and body mass index [BMI]) that were selected based on previous research.3,11 Sample sizes in the development and validation samples reflect the number of patients in each of the samples who met all the abovementioned criteria. A patient's first visit satisfying the inclusion criteria served as their index visit.
Study Measures
The primary outcome was time to LoA. In the development sample, this was defined as the time from the index visit to the first subsequent visit when the patient was unable to complete the 6MWD test, even if the patient was able to complete the test at a later visit. For patients who maintained the ability to complete the 6MWD test during follow-up, time to LoA was censored at the last available assessment. 6MWD and timed function tests were assessed using published methodology16,17 across all data sources comprising the development sample. Inability to complete the 6MWD test was unambiguously recorded as 6MWD = 0 m.
In the validation sample, time to LoA was defined as time to wheelchair dependence. The 6MWD was not systematically available for all CINRG sites, precluding its use as an end point in the validation sample. Because assessment of LoA varies across clinical settings,18 we used time to wheelchair dependence, a clinically meaningful milestone in DMD disease progression and a well-established surrogate for LoA. Wheelchair dependence was defined as patient-reported or caregiver-reported age at continuous wheelchair use, approximated to the nearest month, and verified by a trained clinical evaluator who confirmed LoA based on inability to perform the 10MWR.
Candidate prognostic factors included measures assessed at the index visit across all data sources: age, timed function tests (10MWR, 4SC, RFF), height, weight, and BMI. Timed function test outcomes were truncated at 30 seconds for consistency across data sources. Patients unable to perform the 4SC and/or RFF because of disease progression were given a time of 30 seconds. Patients with missing data on study measures were excluded.
Statistical Methods
Development of the Prognostic Score
Classification and regression tree (CART) models19 of time to LoA were used to classify patients into subgroups with different levels of risk of LoA over time using the candidate prognostic factors as predictors. Fitted models output binary decision trees identifying the most relevant prognostic factors and classification thresholds for these factors, based on which patients were classified into risk groups. To reduce risk of overfitting and increase generalizability,19 restrictions were imposed on the minimum number of patients in each risk group (N = 20) and the minimum number of patients required before the algorithm made a split (N = 50).
Simplification of the Prognostic Score
Simplifications to the originally fitted CART model were explored to see whether classification of patients in clinical settings could be made easier without sacrificing predictive power. A simplified prognostic score was generated by prioritizing consideration of RFF and 10MWR as candidate predictors over 4SC because these are more commonly assessed in routine clinical practice, rounding classification thresholds and combining multiple, similarly performing risk groups. Risk group classifications based on the simplified prognostic score were generated and compared with those from the original score.
Assessment of the Simplified Prognostic Score in the Development Sample
Patient characteristics at the index date were summarized in the development sample (Table 1), among all patients, and by risk group classifications derived from the simplified model. The ability of the simplified prognostic score to differentiate patients at different risks of LoA over time was assessed using Kaplan-Meier (KM) analyses of time to LoA, stratified by risk group. An alpha level of 0.05 was used to assess statistical significance.
Table 1.
Patient Characteristics in the Development Sample at the Index Visit, Among All Patients, and by Risk Group
| All patients N = 608 |
Risk group 1 N = 189 |
Risk group 2 N = 170 |
Risk group 3 N = 141 |
Risk group 4 N = 73 |
Risk group 5 N = 35 |
|
| Demographics | ||||||
| Age (y) | 9.0 ± 2.4 | 7.8 ± 1.7 | 8.6 ± 2.1 | 9.7 ± 2.2 | 10.9 ± 2.4 | 11.5 ± 2.8 |
| Age, categories (y) | ||||||
| <6 | 46 (7.6) | 24 (12.7) | 20 (11.8) | 2 (1.4) | 0 (0.0) | 0 (0.0) |
| 6–≤8 | 182 (29.9) | 88 (46.6) | 54 (31.8) | 32 (22.7) | 6 (8.2) | 2 (5.7) |
| 8–≤10 | 206 (33.9) | 59 (31.2) | 58 (34.1) | 51 (36.2) | 26 (35.6) | 12 (34.3) |
| 10–≤12 | 105 (17.3) | 14 (7.4) | 28 (16.5) | 36 (25.5) | 20 (27.4) | 7 (20.0) |
| 12–≤14 | 48 (7.9) | 4 (2.1) | 7 (4.1) | 15 (10.6) | 14 (19.2) | 8 (22.9) |
| ≥14 | 21 (3.5) | 0 (0.0) | 3 (1.8) | 5 (3.6) | 7 (9.6) | 6 (17.1) |
| Corticosteroids | ||||||
| Corticosteroid use | ||||||
| On prednisone | 257 (42.3) | 67 (35.5) | 76 (44.7) | 73 (51.8) | 24 (32.9) | 17 (48.6) |
| On deflazacort | 351 (57.7) | 122 (64.6) | 94 (55.3) | 68 (48.2) | 49 (67.1) | 18 (51.4) |
| Daily regimen (vs other)a | 361 (76.5) | 109 (87.9) | 110 (78.6) | 76 (64.4) | 41 (70.7) | 25 (78.1) |
| Corticosteroid duration (mo) | ||||||
| <12 | 166 (27.5) | 58 (31.0) | 54 (32.1) | 36 (25.5) | 14 (19.4) | 4 (11.4) |
| 12–24 | 121 (20.1) | 43 (23.0) | 37 (22.0) | 24 (17.0) | 11 (15.3) | 6 (17.1) |
| ≥24 | 316 (52.4) | 86 (46.0) | 77 (45.8) | 81 (57.5) | 47 (65.3) | 25 (71.4) |
| Ambulatory function | ||||||
| 6MWD (m) | 360.1 ± 78.9 | 409.6 ± 60.6 | 385.1 ± 62.3 | 333.6 ± 47.9 | 295.33 ± 61.4 | 213.4 ± 56.9 |
| Timed 10MWR (s) | 6.2 ± 2.6 | 4.3 ± 1.0 | 5.3 ± 1.1 | 7.1 ± 1.3 | 8.3 ± 1.8 | 12.5 ± 4.0 |
| Timed 10MWR, categories | ||||||
| <10 s | 561 (92.3) | 188 (99.5) | 169 (99.4) | 141 (100.0) | 63 (86.3) | 0 (0.0) |
| ≥10 s | 47 (7.7) | 1 (0.5) | 1 (0.6) | 0 (0.0) | 10 (13.7) | 35 (100.0) |
| Timed RFF (s) | 9.8 ± 9.0 | 3.1 ± 0.5 | 5.2 ± 0.8 | 11.1 ± 3.1 | 26.3 ± 5.9 | 28.8 ± 2.9 |
| Timed RFF, categories | ||||||
| <5 s | 264 (43.4) | 189 (100.0) | 75 (44.1) | 0 (0.0) | 0 (0.0) | 0 (0.0) |
| ≥5 s | 344 (56.6) | 0 (0.0) | 95 (55.9) | 141 (100.0) | 73 (100.0) | 35 (100.0) |
| Timed 4SC (s) | 5.6 ± 5.4 | 2.4 ± 0.8 | 3.8 ± 1.6 | 6.2 ± 2.2 | 10.9 ± 6.6 | 18.7 ± 9.0 |
| Timed 4SC, categories | ||||||
| <3 s | 213 (35.0) | 155 (82.0) | 51 (30.0) | 6 (4.3) | 1 (1.4) | 0 (0.0) |
| ≥3 s | 395 (65.0) | 34 (18.0) | 119 (70.0) | 135 (95.7) | 72 (98.6) | 35 (100.0) |
| NSAA total score | 22.7 ± 7.0 | 28.7 ± 3.6 | 25.9 ± 4.7 | 19.1 ± 4.8 | 15.1 ± 3.6 | 10.4 ± 3.1 |
| NSAA total score, categories | ||||||
| <12 | 30 (6.7) | 0 (0.0) | 0 (0.0) | 6 (5.1) | 9 (16.1) | 15 (62.5) |
| 12–≤18 | 80 (17.9) | 1 (0.9) | 3 (2.3) | 39 (33.1) | 28 (50.0) | 9 (37.5) |
| 18–≤24 | 115 (25.7) | 6 (5.1) | 40 (30.1) | 50 (42.4) | 19 (33.9) | 0 (0.0) |
| ≥24 | 223 (49.8) | 110 (94.0) | 90 (67.7) | 23 (19.5) | 0 (0.0) | 0 (0.0) |
| Vitals | ||||||
| Height (cm) | 123.3 ± 11.2 | 116.7 ± 8.5 | 121.2 ± 9.5 | 128.2 ± 10.9 | 131.6 ± 10.9 | 131.8 ± 9.2 |
| Weight (kg) | 29.7 ± 10.4 | 24.6 ± 6.1 | 27.6 ± 8.6 | 33.2 ± 11.3 | 37.4 ± 11.5 | 37.3 ± 11.9 |
| BMI (kg/m2) | 19.1 ± 4.0 | 17.9 ± 2.6 | 18.4 ± 3.8 | 19.8 ± 4.5 | 21.2 ± 4.5 | 21.1 ± 5.0 |
Abbreviations: 4SC = 4-stair climb; 6MWD = 6-minute walk distance; 10MWR = 10-m walk/run; BMI = body mass index; NSAA = North Star Ambulatory Assessment; RFF = rise from floor.
Means and SDs are shown for continuous characteristics; counts and percentages are shown for categorical characteristics.
Risk group definitions: risk group 1: RFF <4 seconds; risk group 2: RFF ≥4 to <7 seconds; risk group 3: RFF ≥7 to <20 seconds and 10MWR <10 seconds; risk group 4: RFF ≥7 to <20 seconds and 10MWR ≥10 seconds OR RFF ≥20 seconds and 10MWR <10 seconds; risk group 5: RFF ≥20 seconds and 10MWR ≥10 seconds.
Data on corticosteroid regimen were not available for all patients. Percentages are among those with available data on corticosteroid regimen.
Predictive performance of the simplified score relative to the original score was also assessed. Risk stratification based on the simplified vs the original score was assessed using KM analyses. The pseudo–R-squared value from a Weibull regression model for time to LoA containing the simplified score as a predictor was compared with the pseudo–R-squared values from models containing (1) only the original score and (2) baseline age, 10MWR, 4SC, RFF, height, weight, and BMI.
External Validation of the Simplified Score
External validity was assessed by applying the simplified score in the validation sample and comparing the KM curves stratified by risk groups between the development and validation sample.
Sensitivity Analyses
To provide further context, classifications of patients based on the simplified score were compared with previously proposed threshold-based classifications,3,17,20 including a single cutoff of 5 seconds for RFF, or thresholds of age ≤7 years and 6MWD ≤350 m. To assess robustness of the classifications to corticosteroid type, risk group classifications were assessed separately for patients treated with deflazacort vs with prednisone at index. Finally, given differences in follow-up times and baseline patient characteristics across the clinical trial and RWD/NHD sources analyzed here, we also investigated whether data sources significantly predicted time to LoA after adjusting for differences in baseline prognostic factors.
Data Availability
All relevant aggregate data are reported within the article and the Supporting Information files. This study uses third-party data sources accessed by the cTAP, through data use agreements with the relevant data holders. Individual-level data for 2 of these data sources (PRO-DMD-01 and tadalafil DMD trial) are available on Vivli, a public repository, and can be accessed through requests to the data holders through Vivli. Other data sources used are not available on public repositories and may be available through data use agreements with the individual data holders. Requests for individual patient data may be directed toward the individual institutions/organizations that have collected, curated, and/or hold these patient data. These organizations will consider data requests according to their own data sharing policies and governance.
Results
Data from 608 boys who met the inclusion criteria were analyzed for development of the prognostic score (Table 1, eTable 2). At the index visit, the mean age was 9.1 (range 4.4–19.4) years and the mean 6MWD was 360 m (SD 79 m). Over half of the sample (53%) was followed for at least 1 year, and 37%, 21%, and 13% of the sample were followed for at least 2, 3, and 4 years, respectively.
Over 1,241 patient-years of follow-up, 116 boys experienced LoA. The median time to LoA from the index date was 4.6 years, and 99%, 91%, and 81% of patients were free of LoA at 6, 12, and 24 months, respectively. Among patients observed to reach LoA, the mean age at time of event was 12.2 (range 6.4–19.7) years.
Assessment of the Original and Simplified Prognostic Scores
Classifications trees based on the original and simplified models are shown in eFigure 1. The original model identified 6 different risk groups based on thresholds of RFF, 10MWR, and 4SC and had a pseudo–R-squared of 0.36 (eFigure 1A). Notably, this measure of fit was higher than that obtained in a proportional hazards model including all candidate predictors as independent variables (0.29). In this model, RFF and 10MWR were the most important predictors and were relevant for classifying patients into all 6 risk groups, while 4SC was relevant only for further distinguishing patients with RFF values less than approximately 7 seconds (risk group 1 or 2). In this original model, patients in risk groups 4 and 5 were found to have similar median times to LoA of 1.5 and 2 years, respectively, and similar proportions of patients remained ambulatory at 6 months (96% vs 100%) and 12 months (77% vs 74%, respectively) after the index date (eFigure 2).
To facilitate easier use in practice, the original model was simplified by rounding split point thresholds, removing 4SC, and combining risk groups 4 and 5 (eFigure 1B). In the simplified model, RFF and 10MWR were identified as predictors and the model had similar prognostic power (pseudo-R2 of 0.33; eFigure 1B). This simplified classification required that patients complete only RFF and 10MWR, with only RFF needed for patients who completed RFF in under 7 seconds.
KM curves for time to LoA for each risk group identified based on the simplified model were well separated and differed in time to LoA (Figure 1). Median time to LoA for risk group 1 was not reached, whereas median times to LoA for groups 2–5 were 4.4 (95% CI 4.0–inf), 3.0 (2.1–4.1), 2.0 (1.6–2.1), and 0.9 (0.8–1.6) years, respectively. A similar relationship was observed for the share of patients experiencing LoA at 6 months, 1 year, 2 years, and beyond (Table 2). Median ages at LoA are summarized in eTable 3.
Figure 1. Kaplan-Meier Analysis of Time to LoA by Risk Group in the Development Sample.
*Statistically significant. Risk group definitions: risk group 1: RFF <4 seconds; risk group 2: RFF ≥4 to <7 seconds; risk group 3: RFF ≥7 to <20 seconds and 10MWR <10 seconds; risk group 4: RFF ≥7 to <20 seconds and 10MWR ≥10 seconds OR RFF ≥20 seconds and 10MWR <10 seconds; risk group 5: RFF ≥20 seconds and 10MWR ≥10 seconds. 10MWR = 10-m walk/run; LoA = loss of ambulation; RFF = rise from floor.
Table 2.
Percentage Risk of LoA Over Time (95% CI), Among All Patients, and by Risk Group
| 6 mo | 1 y | 2 y | 3 y | 4 y | |
| All patients, % | 1.5 (0.5–2.5) | 8.8 (6.1–11.3) | 19.3 (14.9–23.4) | 27.6 (22.0–32.8) | 36.7 (29.2–43.5) |
| By risk group, % | |||||
| Risk group 1 | 0.0 (0.0–0.0) | 0.0 (0.0–0.0) | 1.7 (0.0–4.0) | 1.7 (0.0–4.0) | 6.4 (0.0–12.9) |
| Risk group 2 | 0.0 (0.0–0.0) | 0.0 (0.0–0.0) | 6.4 (0.8–11.8) | 13.0 (3.6–21.4) | 33.1 (11.3–49.5) |
| Risk group 3 | 1.5 (0.0–3.4) | 6.5 (1.4–11.4) | 23.7 (11.0–34.6) | 53.0 (32.4–67.3) | 66.7 (43.1–80.6) |
| Risk group 4 | 1.4 (0.0–4.1) | 29.4 (15.0–41.4) | 61.2 (40.3–74.7) | 81.9 (58.9–92.0) | 87.9 (62.0–96.2) |
| Risk group 5 | 17.1 (3.7–28.7) | 65.8 (40.5–80.4) | 100.0 (—) | 100.0 (—) | 100.0 (—) |
Abbreviations: 10MWR = 10-m walk/run; LoA = loss of ambulation; RFF = rise from floor.
Risk group definitions: risk group 1: RFF <4 seconds (n = 189 at baseline, 10 with LoA over time); risk group 2: RFF ≥4 to <7 seconds (n = 170, 16 with LoA); risk group 3: RFF ≥7 to <20 seconds and 10MWR <10 seconds (n = 141, 32 with LoA); risk group 4: RFF ≥7 to <20 seconds and 10MWR ≥10 seconds OR RFF ≥20 seconds and 10MWR <10 seconds (n = 73, 33 with LoA); risk group 5: RFF ≥20 seconds and 10MWR ≥10 seconds (n = 35, 25 with LoA).
External Validation of the Simplified Prognostic Score
The simplified prognostic score was validated using data from 226 boys in the validation sample who met the inclusion criteria (eTable 4). Over 819 patient-years of follow-up, 73 boys experienced LoA. At the index visit, the mean age was 7.6 ± 2.4 (range 6.6 ± 1.6–11.4 ± 3.3) years. KM curves for each risk group in the validation sample are shown in Figure 2. Risks of LoA, particularly over the first 4 years after the index visit, were similar between the development and validation sample. In the validation sample, the risk groups were well separated, with median times to LoA of not reached for risk group 1, and 7.7, 3.7, 2.2, and 0.6 years for risk groups 2, 3, 4, and 5, respectively. These median times were similar to those in the development sample for risk groups 3, 4, and 5, but longer for risk group 2. The simplified prognostic score was converted into a tabular representation that can be used in clinical practice to classify patients into risk groups (Figure 3).
Figure 2. Kaplan-Meier Analysis of Time to LoA by Risk Groups in the Validation Sample.
Risk group definitions: risk group 1: RFF <4 seconds; risk group 2: RFF ≥4 to <7 seconds; risk group 3: RFF ≥7 to <20 seconds and 10MWR <10 seconds; risk group 4: RFF ≥7 to <20 seconds and 10MWR ≥10 seconds OR RFF ≥20 seconds and 10MWR <10 seconds; risk group 5: RFF ≥20 seconds and 10MWR ≥10 seconds. 10MWR = 10-m walk/run; LoA = loss of ambulation; RFF = rise from floor.
Figure 3. Tabular Representation of the Simplified Prognostic Score for Clinical Use.
10MWR = 10-m walk/run; LoA = loss of ambulation; RFF = rise from floor.
Sensitivity Analyses
The simplified prognostic score provided better risk stratification across the full spectrum of risk than ad hoc thresholds reported in previous research (eFigures 3 and 4). Risk groups identified based on the ad hoc thresholds could be further stratified into subgroups with meaningful differences in LoA risk using the simplified prognostic score. The risk groups based on the simplified score were also well separated by corticosteroid type (prednisone vs deflazacort), as shown in eFigure 5. Duration of corticosteroid use and use of daily corticosteroids at index for risk groups within each corticosteroid type are summarized in eTable 5. In risk groups 1, 3, and 5, time to LoA separation was similar between corticosteroid types, with more separation among patients in risk groups 2 and 4. Finally, in a multivariable Cox model for time to LoA, baseline 6MWD (p < 0.001), baseline rise velocity (p < 0.01), baseline 4SC velocity (p = 0.026), corticosteroid type (p < 0.001), height (p < 0.01), weight (p < 0.01), and BMI (p = 0.047) were all statistically significant predictors, while data source was not (p = 0.12).
Discussion
In this study, we classified patients into 5 risk groups based on 2 widely assessed measures of ambulatory function in the clinic, RFF and 10MWR, to develop a prognostic score for time to LoA in DMD. The risk groups allowed excellent risk stratification of patients, with estimated median times to LoA of approximately 1 year, 2 years, 3 years, 4.5 years, and longer in the development data. The differences across these groups are clinically meaningful and can have profound implications for patient and family/caregiver counseling and planning. The prognostic score was validated using external data, where it performed similarly well and outperformed other previously proposed classifications.3,17,20 Only 2 timed tests are required to apply our prognostic score, facilitating patient evaluation of their prognosis. In patients who can complete the RFF in <7 seconds, the evaluation process is even easier, because the RFF is the only test required to assess the prognostic score. The fact that our prognostic score generates meaningful risk stratification for different definitions of LoA—6MWD = 0 and time to wheelchair dependence—suggests that the identified risk categories are robustly applicable.
All patients included in this study had initiated corticosteroid treatment. Corticosteroid type was not considered as a candidate predictor to inform the prognostic score but analyzed separately. Corticosteroid use has been associated with slowing disease progression in ambulatory patients with DMD.3,21-23 The prognostic meaning of corticosteroid use depends on the efficacy and the shared decision of when to initiate corticosteroids and may not generalize across practices and time; therefore, we studied the time to LoA by risk group among patients treated with deflazacort vs prednisone at index. Findings showed that the greatest differences between patients using deflazacort and prednisone were seen in risk groups 2, 3, and 4. This suggests that progression to LoA may be independent of corticosteroid type in patients who are declining very slowly or very quickly; these findings are consistent with the FOR-DMD trial, which found no significant differences in clinical benefit between patients receiving daily corticosteroid regimens.24
Although age is correlated with functional performance, it did not contribute significant additional prognostic value for time to LoA once multiple measures of baseline function were considered. Approximately 90% of patients included in this analysis were between age 6 and 14 at index, largely reflecting boys expected to have plateauing or declining ambulatory function. Relevance of age as a predictor of time to LoA in younger patients requires further investigation.
Our prognostic score methodology validates progression milestones based on timed function tests. For example, in the PRO-DMD-01 study, 10MWR time >10 seconds (“approaching ambulation” or aLOA) was associated with 100% of patients losing ambulation within 2 years.25 In this study, 100% of patients in risk group 5 lost ambulation within 2 years, supporting the aLOA concept. RFF of <7 seconds was associated with 3 or more years to LoA. Of interest, 100% of patients with RFF >5 seconds lost ambulation within 5 years (eFigure 3). The UK Northstar Clinical Network14 showed that patients with a RFF time of ≤3.5 seconds between 6 and 8 years of age had the mildest disease progression course, and our risk group 1 (RFF <4 seconds) continued ambulating beyond 4 years or more.
We found that median times to LoA were approximately half a year longer in most risk groups in the validation data compared with the development data, possibly because different outcomes were used to assess LoA. In the development sample, assessments of 6MWD occurred every 8–12 weeks in trial placebo arms to every 6 months in RWD/NHD sources. In the validation sample, patient-reported wheelchair dependence, in comparison, may reflect a later event than an assessment of LoA based on 6MWD. While age at loss of ability to perform 10MWR was also available in the validation data, on average, it tended to be later than patient-reported age at full-time wheelchair use. One possible explanation for this is that for most patients, while milestones of wheelchair use and loss of ability to perform 10MWR may both occur between follow-up visits, wheelchair dependence reported by the patient may be recorded as having occurred at an earlier date (between clinic visits), whereas confirmation of inability to perform the 10MWR requires clinical verification at a subsequent visit.
Our goal was to analyze a broad collection of data sources to capture patients with a wide range of baseline ambulatory function and maximize our sample size for the development of a prognostic score to identify patients at different levels of subsequent time to/risk of LoA. As such, the data used in this study come from a diverse collection of sources incorporating both RWD/NHD and clinical trial placebo arm data. RWD/NHD data come from care centers with extensive research and DMD clinical trial experience, with functional assessments usually conducted by physicians, physical therapists, or clinical evaluators with training and experience in administering functional assessments in trial settings. The periods of data collection from these sources preceded the approvals of recent therapies for DMD, and given the globally diverse collection of study sites and multicenter trials included here, standards of care may have been more variable across these sources. However, despite a potential for such differences, we did not find evidence that data source was a significant predictor of time to LoA in our models when considered alongside baseline measures of prognostic function. This suggests that the identified prognostic factors and the developed prognostic score are generalizable to broader ambulatory patient populations receiving corticosteroid standard of care for DMD.
Incorporating our prognostic score for time to LoA may optimize patient care by allowing clinicians to provide more accurate and individualized anticipatory guidance on topics such as need for mobility devices, manual and power wheelchair prescriptions, home modifications for accessibility, financial accommodations, and psychosocial needs,26 providing families time for effective adjustment and care planning. Patient preferences for prognostic information received during counseling can vary from information seeking to avoidance, especially when prognosis does not directly govern evidence-based treatment recommendations.27 Accurate communication of prognostic information, reflecting both expectations and degrees of uncertainty, is challenging even for seasoned practitioners. The developed risk scores are intended to assist care teams; however, it is important to avoid risk of miscommunication or overinterpretation. For example, within each risk group, the median time to LoA could be misinterpreted as a certain or near-certain outcome, when, in reality, patients in each risk group will experience a distribution of times to LoA around that median. Risks of LoA over specific intervals are provided in Table 2 to facilitate proper interpretation. Because risk groups are based on a patient's current function, they can be interpreted as stages of progression from onset of symptoms to LoA.
This prognostic score may be useful in determining patients' eligibility for clinical research. By explaining variation at baseline, prognostic scores can optimize stratification for trial design, inform trial enrichment strategies, and reduce sample size requirements by decreasing unexplained outcome variability.11,12 This prognostic score could also help identify subsets of patients with similar prognosis within a larger, more heterogeneous population or be incorporated into phase 4 disease registries across broader groups of ambulatory patients.
The study has several limitations. Predictions of longer term (e.g., beyond 4 years) risks of LoA are based on limited numbers of patients. An analysis with longer term follow-up data might allow for further stratification of patients within the lowest-risk groups identified here. From a practical perspective, however, classifying LoA over a 2-year horizon may be most clinically informative, as patients will typically follow up with providers at least annually, at which time their prognosis could be updated based on their functional profile. It is unknown whether the prognostic score will be generalizable to a diversity of racial and ethnic non-White populations.
Our prognostic score did not consider other potential prognostic factors, such as corticosteroid regimen, DMD genotypes, other gene modifiers, or MRI biomarkers, because these were not available in all data sources. Previously reported associations between DMD genotype classes amenable to skipping exons 44 and 51 and subsequent times to loss of substantial ambulatory function were smaller than the differences across risk groups identified here.28 It will be important to assess the added prognostic value of MRI-based measures, which have shown longitudinal correlations with function in DMD.29 The average age of participants in the development sample was 9.1 years, and it is plausible that younger boys' prognosis may be more reliant on genetic factors, imaging biomarkers, and other age-appropriate developmental and motor measures relative to older boys.
Conclusion
In summary, this study used a machine learning algorithm on data from natural history databases and placebo arms of clinical trials to develop a simple, easy-to-use prognostic score for time to LoA that can be used in clinical practice. Only knowledge of the performance on 2 timed function tests, RFF and 10MWR, is required. The prognostic score replicated well on a validation data set and performed better than previously proposed classifications used in clinical practice.
Acknowledgment
The authors thank the following contributors of the study: the iMDEX Natural History Study contributors: Joana Domingos (deceased; clinical research fellow; performed study-related activities as clinical research fellow in co-ordinating center) and Valeria Ricotti (study-related activities) (Dubowitz Neuromuscular Centre, University College London Great Ormond Street Institute of Child Health and Great Ormond Street Hospital, London, United Kingdom); Victoria Selby, Amy Wolfe, Lianne Abbott, and Evelin Milev (research physiotherapists; conducted functional assessments), and Efthymia Panagiotopoulou, Mario Iodice, and Maria Ash (study-related activities) (Dubowitz Neuromuscular Centre, Great Ormond Street Hospital for Children NHS Trust, London, United Kingdom); Professor Thomas Voit (Professor of Paediatrics; organized methodological set-up and method and people involvement [Paris site]; [former] Groupe Hospitalier Pitié Salpêtrière, Institut de Myologie, Paris, France; [current] Dubowitz Neuromuscular Centre, University College London Great Ormond Street Institute of Child Health, NIHR Great Ormond Street Hospital Biomedical Research Centre, London, United Kingdom); Valérie Decostre and Stéphanie Gilabert (research physiotherapists; conducted functional assessments), and Jean-Yves Hogrel (site co-investigator; designed and organized functional assessments conducted analyses on strength assessment) (Groupe Hospitalier Pitié Salpêtrière, Institut de Myologie, Paris, France); Alexander Murphy (study-related activities) and Anna Mayhew (research physiotherapists; conducted functional assessments) (John Walton Muscular Dystrophy Research Centre, Newcastle University, Newcastle, United Kingdom); Menno Van der Holst (research physiotherapist, clinical evaluator; carried out functional assessments and study related activities [data entry]; Department of Orthopaedics, Rehabilitation and Physiotherapy, Leiden University Medical Centre, Leiden, the Netherlands); Yvonne D. Krom (research coordinator) and Marjolein J. van Heur-Neuman (research nurse) (study-related activities; Department of Neurology, Leiden University Medical Centre, Leiden, the Netherlands); Merel Jansen and Maaike Pelsma (physiotherapists; delivered functional assessments), and Marian Bobbert (research nurse; study-related activities) (Department of Rehabilitation, Donders Centre of Neuroscience, Radboud University Medical Center, Nijmegen, the Netherlands); Johannes J.G.M. Verschuuren (clinical investigator; designed and organized functional assessments; Leiden University Medical Centre, Leiden, the Netherlands). Cooperative International Neuromuscular Research Group (CINRG) Duchenne Natural History Study (DNHS) contributors: the authors thank the dedicated CINRG DHNS researchers who continue to commit countless hours; the CINRG DHNS network sites that contributed to this project include the following: University of California—Davis-Michelle Cregan, Erica Goude, Merete Glick, Linda Johnson, Jay Han; Holland Bloorview Kids Rehabilitation—Laila Eliasoph, Elizabeth Hosaki, Angela Gonzales, Vivienne Harris; Alberta Children's Hospital—Angela Chiu, Karla Sanchez, Natalia Rincon, Tiffany Haig; Queen Silvia Children's Hospital—Anne-Christine Alhander, Lisa Wahlgren, Anne-Berit Ekstrom, Anna-Karin Kroksmark, Ulrika Sterky; Children's National Medical Center—Marissa Birkmeier, Sarah Kaminski, Allyn Toles; Royal Children's Hospital—Kate Carroll, Katy DeValle, Rachel Kennedy, Andrew Kornberg, Dani Villano; Hadassah Hebrew University Hospital—Adina Bar Leve, Elana Wisband, Debbie Yaffe; Instituto de Neurosciencias Fundacion Favaloro—Luz Andreone, Jose Corderi, Lilia Mesa, Lorena Levi; Children's Hospital of Pittsburgh of UPMC and the University of Pittsburgh—Hoda Abdel-Hamid, Christopher Bise, Ann Craig, Casey Nguyen, Andrea Smith, Jason Weimer; Washington University, St. Louis—M. Al-Lozi, Julaine Florence, Betsy Malkus, Renee Renna, Jeanine Schierbacker, Catherine Seiner, Charlie Wulf; Children's Hospital of Virginia—Susan Blair, Barbara Grillo, Karen Jones, Eugenio Monasterio; University of Tennessee, Memphis—Meegan Barrett-Adair, Judy Clift, Cassandra Feliciano, Rachel Young; Children's Hospital of Westmead—Kristy Rose, Richard Webster, Stephanie Wicks; University of Alberta—Lucia Chen, Cameron Kennedy; The CINRG Coordinating Center—Adrienne Arrieta, Tanisha Brown-Caines, Avital Cnaan, Tina Duong, Fengming Hu, Lauren Morgenroth, Wenze Tang. The authors thank Adina Zhang and Hana Akbarnejad of Analysis Group, Inc., for help preparing tables and figures for this article, and Flora Chik of Analysis Group, Inc., and Mary Smith and Cynthia Snell of SNELL Medical Communication for proofreading, formatting, and supporting the submission of this article.
Glossary
- 4SC
4 stair climb
- 6MWD
6-minute walk distance
- 10MWR
10-m walk/run
- BMI
body mass index
- CART
classification and regression tree
- CINRG
Cooperative International Neuromuscular Research Group
- cTAP
collaborative Trajectory Analysis Project
- DMD
Duchenne muscular dystrophy
- DNHS
Duchenne Natural History Study
- KM
Kaplan-Meier
- LoA
loss of ambulation
- NSAA
North Star Ambulatory Assessment
- RFF
rise from floor
- RWD/NHD
real-world and natural history data
Appendix. Coinvestigators
| Coinvestigators are listed at Neurology.org/N. |
Footnotes
Editorial, page e214882
Author Contributions
C.M. McDonald: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. J. Signorovitch: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. N.M. Goemans: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. E. Mercuri: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. H. Gordish-Dressman: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. K. Vandenborne: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. G. Sajeev: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. S.J. Ward: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. M. Fillbrunn: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. M. Frean: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. L. Servais: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. E.H. Niks: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. V. Straub: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. I.J.M. De Groot: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. F. Muntoni: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data.
Study Funding
This study was conducted by the collaborative Trajectory Analysis Project (cTAP), a precompetitive coalition of academic clinicians, drug developers, and patient foundations formed in 2015 to overcome the challenges of high variation in clinical trials in DMD. CTAP has harmonized data from multiple institutions, registries, and clinical trials across multiple geographies. cTAP has received sponsorship from Astellas (Mitobridge), Avidity Biosciences, BioMarin Pharmaceutical, Bristol Meyers Squibb, Catabasis, Daiichi Sankyo, Dyne Therapeutics, Edgewise Therapeutics, Entrada Therapeutics, FibroGen, Italfarmaco SpA, Marathon Pharmaceuticals, NS Pharma, Pfizer, PTC Therapeutics, Roche, Sarepta Therapeutics, Shire, Solid Biosciences, Summit Therapeutics, Ultragenyx, Vertex Pharmaceuticals, Wave Life Sciences, Parent Project Muscular Dystrophy, Charley's Fund, and CureDuchenne, a founding patient advocacy partner and provider of initial seed funding to cTAP. Physical function testing at Universitaire Ziekenhuizen Leuven was funded by Fonds Spierzieke Kinderen. The PRO-DMD-01 study was sponsored by BioMarin Pharmaceuticals, and data were provided to cTAP by CureDuchenne.
Disclosure
C.M. McDonald has served on advisory Boards for Avidity Therapeutics, Catalyst, Edgewise Therapeutics, Entrada Therapeutics, Roche, Santhera Pharmaceuticals, Sarepta Therapeutics, and Solid Biosciences. He has received consultancy fees from Capricor Therapeutics, Edgewise Therapeutics, FibroGen, Italfarmaco SpA, NS Pharma, PTC Therapeutics, Roche, Santhera Pharmaceuticals, Sarepta Therapeutics, and Solid Biosciences. J. Signorovitch cofounded the cTAP and is an employee of Analysis Group, Inc., a consulting firm that received funding from the membership of cTAP to conduct this study. N. Goemans has served on clinical steering committees and/or as a consultant and received compensation from Eli Lilly, Italfarmaco, PTC Therapeutics, and BioMarin Pharmaceutical and has served as a site investigator for GlaxoSmithKline, Prosensa, BioMarin Pharmaceutical, Italfarmaco, Roche, and Eli Lilly. E. Mercuri has served on clinical steering committees and/or as a consultant for Eli Lilly, Italfarmaco, PTC Therapeutics, Sarepta, Santhera, NS Pharma, Dyne, Entrada, Solid, and Pfizer and has served as PI for GlaxoSmithKline, Prosensa, BioMarin Pharmaceutical, Italfarmaco, Roche, PTC, Pfizer, Sarepta, Santhera, Wave, NS, and Eli Lilly. E. Mercuri is part of an institution that receives funding from Biogen, Roche, and Novartis for a SMA disease registry (ISMAR). H. Gordish-Dressman served as a consultant at Agada Biosciences, Solid BioSciences, Audentes Therapeutics Inc., and TRiNDS LLC. K. Vandenborne has received grants from NIH National Institute of Arthritis and Musculoskeletal and Skin Diseases/National Institute of Neurologic Disorders and Stroke, Parent Project Muscular Dystrophy, and the Muscular Dystrophy Association. She has also received funding from ltalfarmaco SpA, Sarepta Therapeutics, Summit Therapeutics plc, Catabasis Pharmaceuticals, Pfizer Inc., ldera Pharmaceuticals, Bristol-Myers Squibb, and Eli Lilly through grant awards to the University of Florida. G. Sajeev, M Fillbrunn, and M. Frean are current employees of Analysis Group, Inc., a consulting firm that received funding from the membership of cTAP to conduct this study. S.J. Ward cofounded and manages the collaborative Trajectory Analysis Project and has received funding from the membership of cTAP to facilitate this study. L. Servais is a member of the SAB or has performed consultancy for Sarepta, Dynacure, Santhera, Pfizer, Dyne, Entrada, Wave, Sysnav, PTC, RegenexBio, Biomarin, Novartis, Biogen, Cytokinetics, Roche, Audentes Therapeutics, and Affinia Therapeutics. L. Servais has given lectures and has served as a consultant for Roche, Biogen, Avexis, and Cytokinetics. L. Servais is the project leader of the newborn screening in Southern Belgium funded by Avexis, Roche, and Biogen. E.H. Niks is a member of the European Reference Network for Rare Neuromuscular Diseases (ERN EURO‐NMD). E.H. Niks reports grants from Duchenne Parent Project, ZonMW, and AFM; consultancies for BioMarin and Summit; and has served as a local investigator in clinical trials of BioMarin, GSK, Lilly, Santhera, Givinostat, and Roche outside the submitted work. E.H. Niks reports ad hoc consultancies for WAVE, Santhera, Regenxbio, and PTC, and he worked as an investigator of clinical trials of Italfarmaco, NS Pharma, Reveragen, Roche, WAVE, and Sarepta outside the submitted work. V. Straub has participated in advisory boards for Audentes Therapeutics, Biogen, Exonics Therapeutics, Italfarmaco S.p.A., Roche, Sanofi Genzyme, Sarepta Therapeutics, Summit Therapeutics, UCB, and Wave Therapeutics. He has research collaborations with Ultragenyx and Sanofi Genzyme. I.J.M. de Groot has no disclosures. F. Muntoni has received consultancy fees from PTC, Sarepta Therapeutics, Roche, Dyne Therapeutics, Avidity, and Edgewise. The iMDEX natural history study was supported by AFM and coordinated by F. Muntoni at UCL. Go to Neurology.org/N for full disclosures.
References
- 1.Bushby K, Finkel R, Birnkrant DJ, et al. Diagnosis and management of Duchenne muscular dystrophy, part 1: diagnosis, and pharmacological and psychosocial management. Lancet Neurol. 2010;9(1):77-93. doi: 10.1016/S1474-4422(09)70271-6 [DOI] [PubMed] [Google Scholar]
- 2.Ryder S, Leadley RM, Armstrong N, et al. The burden, epidemiology, costs and treatment for Duchenne muscular dystrophy: an evidence review. Orphanet J Rare Dis. 2017;12(1):79. doi: 10.1186/s13023-017-0631-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.McDonald CM, Henricson EK, Abresch RT, et al. Long-term effects of glucocorticoids on function, quality of life, and survival in patients with Duchenne muscular dystrophy: a prospective cohort study. Lancet. 2018;391(10119):451-461. doi: 10.1016/S0140-6736(17)32160-8 [DOI] [PubMed] [Google Scholar]
- 4.Marden JR, Freimark J, Yao Z, Signorovitch J, Tian C, Wong BL. Real-world outcomes of long-term prednisone and deflazacort use in patients with Duchenne muscular dystrophy: experience at a single, large care center. J Comp Eff Res. 2020;9(3):177-189. doi: 10.2217/cer-2019-0170 [DOI] [PubMed] [Google Scholar]
- 5.Dang UJ, Damsker JM, Guglieri M, et al. Efficacy and safety of vamorolone over 48 weeks in boys with Duchenne muscular dystrophy: a randomized controlled trial. Neurology. 2024;102(5):e208112. doi: 10.1212/WNL.0000000000208112 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Birnkrant DJ, Bushby K, Bann CM, et al. Diagnosis and management of Duchenne muscular dystrophy, part 2: respiratory, cardiac, bone health, and orthopaedic management. Lancet Neurol. 2018;17(4):347-361. doi: 10.1016/S1474-4422(18)30025-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Filonova G, Aartsma-Rus A. Next steps for the optimization of exon therapy for Duchenne muscular dystrophy. Expert Opin Biol Ther. 2023;23(2):133-143. doi: 10.1080/14712598.2023.2169070 [DOI] [PubMed] [Google Scholar]
- 8.Mercuri E, Bonnemann CG, Muntoni F. Muscular dystrophies. Lancet. 2019;394(10213):2025-2038. doi: 10.1016/S0140-6736(19)32910-1 [DOI] [PubMed] [Google Scholar]
- 9.Szabo SM, Salhany RM, Deighton A, Harwood M, Mah J, Gooch KL. The clinical course of Duchenne muscular dystrophy in the corticosteroid treatment era: a systematic literature review. Orphanet J Rare Dis. 2021;16(1):237. doi: 10.1186/s13023-021-01862-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Weber FJ, Latshang TD, Blum MR, Kohler M, Wertli MM. Prognostic factors, disease course, and treatment efficacy in Duchenne muscular dystrophy: a systematic review and meta-analysis. Muscle Nerve. 2022;66(4):462-470. doi: 10.1002/mus.27682 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Goemans N, Vanden Hauwe M, Signorovitch J, Swallow E, Song J; Collaborative Trajectory Analysis Project (cTAP). Individualized prediction of changes in 6-minute walk distance for patients with Duchenne muscular dystrophy. PLoS One. 2016;11(10):e0164684. doi: 10.1371/journal.pone.0164684 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Goemans N, Wong B, Van den Hauwe M, et al. Prognostic factors for changes in the timed 4-stair climb in patients with Duchenne muscular dystrophy, and implications for measuring drug efficacy: a multi-institutional collaboration. PLoS One. 2020;15(6):e0232870. doi: 10.1371/journal.pone.0232870 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Muntoni F, Domingos J, Manzur AY, et al. Categorising trajectories and individual item changes of the North Star Ambulatory Assessment in patients with Duchenne muscular dystrophy. PLoS One. 2019;14(9):e0221097. doi: 10.1371/journal.pone.0221097 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Zambon AA, Ayyar Gupta V, Ridout D, et al. Peak functional ability and age at loss of ambulation in Duchenne muscular dystrophy. Dev Med Child Neurol. 2022;64(8):979-988. doi: 10.1111/dmcn.15176 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Stimpson G, Ridout D, Wolfe A, et al. Quantifying variability in motor function in Duchenne muscular dystrophy: UK centiles for the NorthStar ambulatory assessment, 10 m walk run velocity and rise from floor velocity in GC treated boys. J Neuromuscul Dis. 2024;11(1):153-166. doi: 10.3233/JND-230159 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.McDonald CM, Henricson EK, Han JJ, et al. The 6-minute walk test as a new outcome measure in Duchenne muscular dystrophy. Muscle Nerve. 2010;41(4):500-510. doi: 10.1002/mus.21544 [DOI] [PubMed] [Google Scholar]
- 17.Mcdonald CM, Henricson EK, Abresch RT, et al. The 6‐minute walk test and other endpoints in Duchenne muscular dystrophy: longitudinal natural history observations over 48 weeks from a multicenter study. Muscle Nerve. 2013;48(3):343-356. doi: 10.1002/mus.23902 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Camino E, Freed M, Schrader R, et al. 337P Functional assessments used in routine care for Duchenne muscular dystrophy in the USA: a survey of Certified Duchenne Care Centers. Neuromuscul Disord. 2024;43:104441.668. doi: 10.1016/j.nmd.2024.07.677 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Berk RA. Classification and Regression Trees (CART): Statistical Learning From a Regression Perspective. Springer; 2008. [Google Scholar]
- 20.McDonald CM, Riebling P, Souza M, et al. Use of a ≥ 5-second threshold in baseline time to stand from supine to predict disease progression in Duchenne muscular dystrophy. Eur J Paediatr Neurol. 2017;21(1):E237. doi: 10.1016/j.ejpn.2017.04.1267 [DOI] [Google Scholar]
- 21.Kim S, Campbell KA, Fox DJ, Matthews DJ, Valdez R; MD STARnet. Corticosteroid treatments in males with Duchenne muscular dystrophy: treatment duration and time to loss of ambulation. J Child Neurol. 2015;30(10):1275-1280. doi: 10.1177/0883073814558120 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Griggs RC, Miller JP, Greenberg CR, et al. Efficacy and safety of deflazacort vs prednisone and placebo for Duchenne muscular dystrophy. Neurology. 2016;87(20):2123-2131. doi: 10.1212/WNL.0000000000003217 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ricotti V, Ridout DA, Pane M, et al. The NorthStar Ambulatory Assessment in Duchenne muscular dystrophy: considerations for the design of clinical trials. J Neurol Neurosurg Psychiatry. 2016;87(2):149-155. doi: 10.1136/jnnp-2014-309405 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Guglieri M, Bushby K, McDermott MP, et al. Effect of different corticosteroid dosing regimens on clinical outcomes in boys with Duchenne muscular dystrophy: a randomized clinical trial. JAMA. 2022;327(15):1456-1468. doi: 10.1001/jama.2022.4315 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.McDonald CM, Signorovitch J, Mercuri E, et al. Functional trajectories before and after loss of ambulation in Duchenne muscular dystrophy and implications for clinical trials. PLoS One. 2024;19(6):e0304099. doi: 10.1371/journal.pone.0304099 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Porteous D, Davies B, English C, Atkinson J. An integrative review exploring psycho-social impacts and therapeutic interventions for parent caregivers of young people living with Duchenne's muscular dystrophy. Children (Basel). 2021;8(3):212. doi: 10.3390/children8030212 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Crossnohere NL, Armstrong N, Fischer R, Bridges JFP. Diagnostic experiences of Duchenne families and their preferences for newborn screening: a mixed-methods study. Am J Med Genet C Semin Med Genet. 2022;190(2):169-177. doi: 10.1002/ajmg.c.31992 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.McDonald CM, Mayer OH, Hor KN, et al. Functional and clinical outcomes associated with steroid treatment among non-ambulatory patients with Duchenne muscular dystrophy. J Neuromuscul Dis. 2023;10(1):67-79. doi: 10.3233/JND-221575 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Barnard AM, Willcocks RJ, Triplett WT, et al. MR biomarkers predict clinical function in Duchenne muscular dystrophy. Neurology. 2020;94(9):e897-e909. doi: 10.1212/WNL.0000000000009012 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
All relevant aggregate data are reported within the article and the Supporting Information files. This study uses third-party data sources accessed by the cTAP, through data use agreements with the relevant data holders. Individual-level data for 2 of these data sources (PRO-DMD-01 and tadalafil DMD trial) are available on Vivli, a public repository, and can be accessed through requests to the data holders through Vivli. Other data sources used are not available on public repositories and may be available through data use agreements with the individual data holders. Requests for individual patient data may be directed toward the individual institutions/organizations that have collected, curated, and/or hold these patient data. These organizations will consider data requests according to their own data sharing policies and governance.