Characterizing the Occurrence of Key Clinical Milestones in Duchenne Muscular Dystrophy in the United States Using Real-World Data

INTRODUCTION

Duchenne muscular dystrophy (DMD) is a severe X-linked progressive neuromuscular degeneration caused by mutations in the gene for dystrophin, a protein required for the structural integrity of muscle cells [1–4]. Affected patients typically present in early childhood with gait abnormalities, muscle weakness, and delayed motor and cognitive function [1, 5–8]. Several well-described clinical cohorts in the United States (US) have documented the inexorable progression of DMD, including from the Cooperative International Neuromuscular Research Group (CINRG), [9] Duchenne Registry, [10] and Centers for Disease Control and Prevention’s Muscular Dystrophy Surveillance, Tracking, and Research Network (MD STARnet) [11]. Although there is heterogeneity in the exact timing of events, patients inevitably experience key clinical milestones in their progression. Muscular weakness leads to loss of ambulation (LOA), wheelchair use, and scoliosis in late childhood [12–14]. Loss of strength in active breathing muscles contributes to respiratory insufficiency and the need for ventilation in the teenage years [15]. Cardiomyopathy also develops in late adolescence and the progression of DMD culminates with early mortality in the third or fourth decade of life [13, 16–19].

While estimates of the frequency and approximate timing of key clinical milestones are available from these well-defined US clinical cohorts, [9–11] estimates from generalizable population-based real-world cohorts to characterize the progression experienced by those with DMD in routine clinical practice are limited. Administrative claims studies are based on healthcare usage data that document patient diagnoses, procedures performed, medications dispensed, physicians visited, and inpatient stays. Such data are often used to characterize patient populations, understand treatment patterns, and assess aspects of the clinical or economic burden of a health condition from a large population perspective [20]. Claims database assessments for DMD however are limited. At the time of this study, the only published claims-based study providing estimates of the clinical burden of DMD was prior to the widespread use of corticosteroids, and included a small number of patients from one commercial plan only [21]. The objective of this study was to estimate the prevalence of key clinical milestones by age among commercially-insured patients with DMD in the US using real-world data.

MATERIALS AND METHODS

This study was a US-based, real-world, retrospective cohort study to describe the clinical course of patients with DMD, treated under commercial health insurance plans. The study cohort and entry criteria were described previously in a study examining the characteristics of and economic burden among those with DMD in the US [22].

RESULTS

Patient characteristics and health status

Table 1 summarizes patient characteristics and health status over the study period. In total, 1,964 patients with DMD were identified in the primary analysis. Median (IQR) age at cohort entry was 15 (9–21) years, with a median of 1.7 years of follow-up. At least one corticosteroid dispensation was observed among 38.8% over available follow-up; additional breakdown of corticosteroid use by age group at cohort entry is included in the appendix (Appendix Figure A.2). Over the study period, the median (IQR) unweighted Elixhauser score was 1 (0–3), and cardiac and pulmonary comorbidities were the most frequent individual Elixhauser comorbidities observed. Of the other key comorbidities, respiratory infectious disease was the most common (observed in 48.8% of the cohort) followed by anxiety disorders (15.4%) and asthma (14.5%).

The prevalence of of key clinical milestones

Among the DMD cohort from the primary analysis, neurologic and/or neuropsychiatric involvement was observed among 27%, wheelchair use among 45%, scoliosis among 30%, cardiomyopathy among 46%, and respiratory involvement among 34%. The ages at first observed claim for each key clinical milestone in this cohort are presented in Fig. 1. The median (IQR) age at neurologic and/or neuropsychiatric involvement was first recorded at 11 (8–16) years old; wheelchair use was 16 (12–22) years old; for scoliosis, 16 (12–20) years old; for cardiomyopathy, 17 (13–22) years old; and respiratory involvement, 19 (15–24) years old. Age-restricted analyses were also performed to estimate the occurrence of those key outcomes stratified by age at cohort entry. Compared to the overall cohort, the prevalence of key clinical milestones was higher among the age- and time-restricted subset (Table 2), as expected given that the age and time restrictions aimed to focus the observation window on the time period where these outcomes were more likely to have first occurred. As the higher observed prevalence could also be due, in part, to the minimum follow-up requirement, a variation of this analysis was explored where the minimum follow-up requirement was removed (sensitivity analysis #1), and varied to explore the impact of a shorter observation window (sensitivity analysis #2). Due to the nature of age-restriction, the median age at outcome was largely correlated with the respective age restriction across all scenarios.

Fig. 1

Age at first record of key clinical milestones, and the number and percentage observed with each, among patients with DMD. *N/N involvement: neurologic/neuropsychiatric involvement; Box represents the interquartile range (IQR), from the first quartile (Q1:25%) to the third quartile (Q3:75%), with the line in the middle depicting the median; dots represent outliers, defined as greater than the range of Q1 - 1.5×the interquartile range to Q3 + 1.5×the IQR, as bounded by the extended lines.

The age-restricted KM analyses showed that as follow-up (time from cohort entry) increases, the prevalence of events captured increases (Table 3). By year one, approximately 25% of patients aged 8 to 10 years at cohort entry had an event of wheelchair use captured, compared to 40% by year two, and over 50% by year three. The corresponding values were higher among those aged 11 to 13 years, ranging from approximately 40% by year one to 60% by year three. Similar trends are observed for other key clinical milestones, with a few showing trends of plateau between year two and three, such as respiratory involvement among those aged 17 to 19 years at index which started to plateau at approximately 42% by the end of year two.

Additional sensitivity analyses

In the sensitivity analysis where the criterion for a minimum of one-year of continuous follow-up was removed, the observed percentage experiencing key clinical milestones was less than in the age-restricted analyses. However, estimates generally remained higher than in the primary analyses (Table 2). In addition, in the sensitivity analysis where the observation window was shortened to one year, the observed percentage experiencing key clinical milestones was expectedly even lower than other scenarios explored.

The sensitivity analyses exploring different DMD cohort definitions (i.e. restricting to patients < 18 years old at index, requiring patients to have at least one diagnosis from a specialist, or excluding patients with myoneural disorder, Guillain-Barre syndrome, or hereditary motor and sensory neuropathy) showed that patients identified by these definitions had similar durations of follow-up, and frequency of corticosteroid use compared to the primary analyses cohort. Age at index, by definition, was lower among the cohort < 18 years old at index, however remained comparable across the other sensitivity analysis cohorts. The frequency and age at key clinical milestones were also comparable between the sensitivity and primary analyses, with the exception of the cohort who was < 18 years old at index, which by definition would have lower estimated median ages at the occurrence of these comorbidities due to the elimination of older DMD patients. Among the cohort < 18 years old at index, the overall frequency of patients with a record of LOA was comparable to that of the base case definition; median ages were higher for outcomes occurring earlier in patients’ lives, and lower for outcomes that tend to occur in the later stages (Appendix Table A.3).

DISCUSSION

Large well-conducted clinical studies and registries have documented the characteristics of the clinical progression of DMD; [12, 13, 18, 19, 39] but how these align with estimates based on real-world data has not previously been reported. This health insurance claims study used the large representative MarketScan commercial database to estimate the prevalence of notable clinical milestones that characterize the natural history of DMD, by age. Approximately 2,000 patients with commercial insurance coverage in the US were included.

The results of the primary (non-age-restricted) analyses showed a slightly older median age at key clinical milestones compared with published estimates from clinical studies, [12, 13, 18, 19, 39] which suggested that at least for some patients, the events captured were potentially not the initial diagnosis related to that milestone. However, as is expected given the relatively short follow-up available in the dataset, the observed frequency of patients experiencing each milestone was less than anticipated overall. As a result, age-restricted analyses were performed. These analyses generally showed higher frequencies of key clinical milestones compared to the primary analysis, as expected given that the age and time restrictions aimed to focus the observation window on the time period where these outcomes were more likely to have occurred. The frequency of wheelchair use, as a proxy for LOA, among children and young teenagers in the present study – observed in 55% – was consistent with published estimates of the age at wheelchair use range from 30% by 10 years [12] through 95% by 15 years of age [13]. The frequency of scoliosis, documented in up to 50% of young teenagers over their first two years of follow up, was also consistent with published estimates suggesting 60% of DMD patients have scoliosis by 15 years of age [13]. Neurologic and/or neuropsychiatric involvement was observed in less than half of children and young teenagers in the current study, which is also broadly consistent with published estimates of the age at diagnosis of common neuropsychiatric complications in DMD; [40] unlike LOA, for example, it is not necessarily to be expected that these complications would be identified in everyone with DMD [41]. Respiratory involvement was observed in almost half of the cohort of older teenagers, and published estimates suggest 40% to 50% of patients require ventilation over 20 years of age [39, 42]. Finally, estimates of cardiomyopathy and heart failure were on the lower end of published estimates of 68% to 93% diagnosed with cardiomyopathy by 20 years of age [13, 18, 19].

The potential challenges in ascertaining clinical outcomes when using administrative data warrant further discussion. The MarketScan databases are based on claims submitted for reimbursement and not research purposes, and billing practices vary between providers and by insurance type. In the current study, identification of scoliosis, neurologic involvement, cardiomyopathy, respiratory failure, and heart failure primarily relied on diagnosis codes. However, the presence of a diagnostic code alone cannot indicate the severity of the underlying condition; nor are the underlying reasons for a physician selecting a particular diagnostic code available. For example, whether the use of a specific code indicates a clinical suspicion of early signs of a complication versus a severe manifestation, or whether a medication is used prophylactically or for acute treatment, is not clear. Some outcomes may not be easily identified in administrative data, particularly capture of initial symptoms or behaviors. For example, some of the diagnoses contributing to the neuropsychiatric/neurologic involvement outcome are challenging to identify using claims datasets [43]. Of necessity, selected outcomes in the current study were ascertained based on proxy measures within the claims datasets and the reliability of these is unclear [44]. For LOA, as a specific diagnostic code is not available, assessment was based on wheelchair use. However, identification of a wheelchair code does not indicate if it is a first purchase or a replacement; or give information of the frequency of wheelchair use. Equally, wheelchairs may be acquired via other sources. In a similar vein, for identifying respiratory involvement, ventilation use may indicate the start of respiratory decline or a management strategy for more severe disease. Finally, across all the outcomes considered, it is possible that a given encounter was coded as a visit for ‘DMD’, rather than for the specific manifestation of interest. Despite these challenges, these data provide indicators of how patients progress when managed in a standard clinical practice setting and have implications for using real-world data to monitor the clinical burden of patients with DMD over time.

There are some additional limitations to the analyses that should be noted. A clinically validated case definition for use in administrative claims databases is not available and there may have been some misclassification on exposure (e.g. patients with other MDs including limb-girdle or Becker muscular dystrophy may have been inadvertently included in the study cohort and patients with DMD over age 30 may have been inadvertently excluded). Other methods to help eliminate other MD cases, for example restricting the timing of the first observed MD diagnosis to early childhood, were not implemented because of the relatively short follow-up per patient available to allow true ascertainment of the first diagnosis, and also to avoid 1) a prohibitively small sample size for analysis, and 2) disregarding true cases of DMD who entered the cohort at older ages (due to the limited follow-up available within MarketScan). Nevertheless, additional sensitivity analyses performed to explore the impact of case definition assumptions showed little variability in estimates of study outcomes. In addition, while the sensitivity analysis restricting to those < 18 years at index led to lower median age estimates, the direction of variation in the frequency of key clinical outcomes observed supported the base case findings on the prevalence of these events. Wheelchair use was consistently documented across sensitivity analysis cohorts; and events occurring earlier in life were observed at a higher frequency (while events occurring later in life were observed at a lower frequency) among the cohort restricted to those < 18 years at index. One additional limitation to note with regard to the definition of DMD, is that individuals with DMD who appear in the MarketScan databases over age 30 years for the first time would be missed; however, this would likely occur quite rarely.

Given the limited follow-up available per patient and that the age at cohort entry of the overall population was in the teenage years, the timing of key clinical milestones occurring prior to cohort entry cannot be accurately ascertained. It therefore cannot be determined whether the first event observed was the first occurrence of the key milestone in a patient’s life (rather than an acknowledgement of the patient having already passed that milestone or ongoing documentation of what has become a chronic condition). As exemplified through the age-restricted sensitivity analyses varying follow-up and observation window requirements, follow-up duration plays a large role in the capture of key milestones in a patient’s life. As a result, this study presents the prevalence of key milestones by age at cohort entry and fixed observation windows to allow better interpretation of findings. To supplement this, the KM analyses showed that the prevalence of events captured increases as follow-up increases.

Additionally, a challenge in comparing the occurrence of outcomes with earlier studies is introduced by developments in research and diagnostics such that certain complications (e.g. early signs of cardiomyopathy) may be identified at a younger age through the increased sensitivity of tools such as cardiac MRI. This may therefore lead to apparent higher rates and earlier identification compared to earlier studies with outcomes based on older technology. While understanding mortality is of interest, very limited data were available (i.e. only from inpatient records, and only from the beginning of the study period through 2016); this outcome was not investigated further. Assessing corticosteroid treatment patterns was not a primary objective of the study and as such it was not designed to comprehensively assess these. As with the analyses of key clinical milestones, it is important to consider the window of follow-up data availability when interpreting available data on corticosteroid use [45]. Given that the majority of this cohort was outside of the age range with the highest anticipated corticosteroid use (6 to 12 years of age), and a large proportion of patients had lost ambulation during the study, the lower rate of overall corticosteroid use in this study cohort is expected and should be interpreted in this context. Finally, the results of this study are specific to individuals covered under a subset of commercial plans and may not be generalizable to all commercial or to government payer segments in the US.

The use of well-validated datasets that provided a large sample size of DMD patients, including children and young adults, from varied commercial insurance plans, was a key strength of this study. Performing age- and time-restricted analyses helped to address potential biases associated with estimating time-dependent outcomes from individuals of differing ages over the course of a non-standardized follow-up window. However, it is important to remember that there is between-patient heterogeneity in the timing of the occurrence of these outcomes, [14, 16] which complicates accurate ascertainment when considered in the context of relatively short follow-up windows per patient available in claims datasets. The results of the sensitivity analyses for the cohort definition were consistent with those of the primary analyses, providing support that the study findings were robust. With respect to the cohort definition, identifying patients with DMD-genotype specific treatments (such as eteplirsen) or DMD/BMD-specific ICD-10 codes (G71.01) was also explored, but did not identify any additional patients beyond those already captured by the study inclusion criteria.

CONCLUSIONS

This is the first study to document the occurrence of key clinical milestones among real-world patients with DMD with a wide range of commercial insurance plans. The prevalence of key clinical milestones observed by age was broadly consistent with published findings from large registries. Variability in estimates reflect the clinical heterogeneity experienced by those with DMD but also the impact of observation windows on the ability to ascertain key clinical outcomes in DMD. As such, the findings of this study also help delineate the types of outcomes that one can well characterize using existing claims data for cohort studies, and considerations on the methodology to do so. These data summarizing the occurrence of relevant clinical outcomes among patients with DMD with commercial insurance coverage add to the growing body of evidence describing the clinical course of DMD patients using real-world data.

DISCLOSURES

FUNDING

This work was supported by Sarepta Therapeutics, Inc.

CONFLICT OF INTEREST

SMS, CQ and EP are employees of Broadstreet HEOR, which received funds from Sarepta Therapeutics Inc for this work.

ACK and KLG are employees of Sarepta Therapeutics, Inc.

SI has received research funding or consulting fees from Avexis, Biogen, Fibrogen, Mallinkrodt, Regeneron, Sarepta, Scholar Rock, PTC Therapeutics, Pfizer, MDA, CureSMA, NIH, Genentech-Roche, and BCBS.