Rate of Change in Cardiac Magnetic Resonance Imaging Measures Is Associated With Death in Duchenne Muscular Dystrophy

Abstract

Background

Cardiovascular disease is the leading cause of death among patients with Duchenne muscular dystrophy (DMD). Identifying patients at risk of early death could allow for increased monitoring and more intensive therapy. Measures that associate with death could serve as surrogate outcomes in clinical trials.

Methods and Results

Duchenne muscular dystrophy subjects prospectively enrolled in observational studies were included. Models using generalized least squares were used to assess the difference of cardiac magnetic resonance measurements between deceased and alive subjects. A total of 63 participants underwent multiple cardiac magnetic resonance imaging and were included in the analyses. Twelve subjects (19.1%) died over a median follow‐up of 5 years (interquartile range, 3.1–7.0). Rate of decline in left ventricular ejection fraction was faster in deceased than alive subjects (P<0.0001). Rate of increase in indexed left ventricular end‐diastolic (P=0.0132) and systolic (P<0.0001) volumes were higher in deceased subjects. Faster worsening in midcircumferential strain was seen in deceased subjects (P=0.049) while no difference in global circumferential strain was seen. The rate of increase in late gadolinium enhancement, base T1, and mid T1 did not differ between groups.

Conclusions

Duchenne muscular dystrophy death is associated with the rate of change in left ventricular ejection fraction, midcircumferential strain, and ventricular volumes. Aggressive medical therapy to decrease the rate of progression may improve the mortality rate in this population. A decrease in the rate of progression may serve as a valid surrogate outcome for therapeutic trials.

Nonstandard Abbreviations and Acronyms

DMD

Duchenne muscular dystrophy

Ecc

circumferential strain

FWHM

full‐width half‐maximum

LGE

late gadolinium enhancement

Clinical Perspective.

What Is New?

• More rapid rate of change in cardiac magnetic resonance imaging measures over time is associated with early death in Duchenne muscular dystrophy.

• Rate of change in left ventricular ejection fraction, volumes, and strain associated with death.

What Are the Clinical Implications?

• Clinical trials for Duchenne muscular dystrophy cardiomyopathy are difficult to perform, as the intervention must occur long before expected death, but the Food and Drug Administration has signaled a willingness to accept surrogate outcome measures.

• Cardiac magnetic resonance imaging measures may identify patients likely to have early death and provide surrogate outcomes for clinical trials.

• Larger, multicenter studies are needed to allow for improved generalizability, multivariable modeling, and better characterization of trends among older patients.

Duchenne muscular dystrophy (DMD) is an X‐linked recessive disorder caused by mutations in the dystrophin gene leading to progressive muscle degeneration, respiratory failure, and cardiomyopathy. DMD affects ≈20 in every 100 000 live male births. ¹ Life expectancy has gradually increased with improved ventilatory techniques and technologies, and cardiovascular disease and cardiomyopathy are now the primary cause of death. ² , ³ , ⁴ Identifying patients at risk of progressive cardiomyopathy and early death could allow for more intensive, targeted therapeutic management. Cardiomyopathy develops with median onset at about 14 years of age, ⁵ while death more often now occurs in the third or fourth decade of life. ⁴ This makes clinical trials difficult, as the time gap between the ideal initiation of medical therapy and death makes trials using death as the primary outcome prohibitively long. This has led to the search for surrogate outcomes that associate with death both to guide clinical management and for use in clinical trials.

Previous imaging studies have aimed to identify parameters associated with death. Most studies have analyzed the association of echocardiographic indices, including measures of function and chamber size. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ However, echocardiography has significant limitations due to poor acoustic windows. Fewer studies have investigated the association of cardiac magnetic resonance (CMR) measures with death. Late gadolinium enhancement (LGE) on CMR has previously been associated with death ¹¹ and adverse events ¹² in small studies. Left ventricular ejection fraction (LVEF) measured by CMR has similarly been associated with death. ⁹ , ¹¹ Our recent work has shown that LGE and LVEF can be combined into an imaging risk score that predicts death. ¹³ Recent work within our patient cohort has also shown that additional CMR measures, including indexed left ventricular end‐diastolic (LVEDVi) and end‐systolic (LVESVi) volumes as well as circumferential strain (Ecc) associate with death. ¹⁴

Intuitively, the rate of change of these imaging metrics would also associate with death. The only prior study to evaluate rate of change by echocardiography demonstrated increased death in patients with onset of left ventricular (LV) dysfunction before 18 years of age but no difference in the rate of change of LV function between groups. ¹⁰ If imaging measures are to be used as surrogate end points in management and clinical trials, the rate of change in these metrics would ideally be associated with death, as modification in this rate of change would be the end point of such studies. Moreover, assessment of preceding rates of change may allow for risk stratification of patients before starting a clinical trial. No studies have investigated this association. We aimed to determine the rate of change in CMR measures over time and determine whether this rate of change is associated with death.

Methods

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Ethical Considerations

This study was approved by the Institutional Review Board at Vanderbilt University Medical Center. Written informed consent was obtained for all participants in the underlying observational studies, and enrollment was completed between January 2013 and January 2020. Participants aged <18 years signed an age‐appropriate assent form. Two of the authors (J.H.S. and J.R.S.) had access to all the data in the study and take responsibility for its integrity and the data analysis.

Study Population

DMD diagnosis was confirmed by skeletal muscle biopsy or presence of a dystrophin mutation in the setting of a clinical phenotype of muscle weakness. Study participants were predominantly from patients with DMD enrolled in ongoing prospective observational studies that include longitudinal CMR (N=72). A smaller number of patients who had consented for a CMR with research sequences were also included (N=6). All CMR images for included participants were used for the study, including those obtained before enrollment in prospective studies to increase power, making this an ambispective study. As the focus of this study was the longitudinal change in CMR metrics, only subjects with >1 CMR were included in the analyses (N=63).

CMR Protocol

CMR was performed on a 1.5 Tesla Siemens Avanto (Siemens Healthcare Sector, Erlangen, Germany) with an 8‐channel cardiac coil or a 1.5 Tesla Siemens Avanto Fit (Siemens) with a 32‐channel coil. CMR protocol was performed as previously described. ¹⁵ In brief, CMR protocol included functional imaging using balanced steady‐state free precession imaging. ¹⁶ Myocardial tagging was performed in the short axis at the base, level of the papillary muscles, and apex using a segmented k‐space fast gradient echo sequence with ECG triggering as previously described. ¹⁷

Intravenous gadolinium contrast (gadopentate dimeglumine, Magnevist; Bayer Healthcare Pharmaceuticals, Wayne, NJ, at a dose of 0.2 mmol/kg; or gadobutrol, Gadavist, Bayer Healthcare Pharmaceuticals, at a dose of 0.15–0.2 mmol/kg) was administered through a peripheral intravenous line. LGE was performed using single‐shot and segmented inversion recovery (optimized inversion time to null myocardium) and phase‐sensitive inversion recovery (inversion time of 300 milliseconds) imaging in the 4‐chamber, 3‐chamber, and 2‐chamber planes as well as the short‐axis stack.

Breath‐held modified Look‐Locker inversion recovery sequences were performed as previously described, ¹⁵ before and 15 minutes after contrast administration at the base, midventricular level, and apex in the short‐axis plane at the same slice location as the T2‐mapping and tagged images. ¹⁸ , ¹⁹ Modified Look‐Locker inversion recovery sequences were ECG‐triggered images obtained in diastole with precontrast acquired as a 5(3s)3, and postcontrast protocol was acquired at a 4(1)3(1)2. ²⁰ Motion correction was performed, and a T1 map was generated on the scanner. ²¹ Any image felt to be inadequate due to poor breath holds or poor motion correction was repeated at the time of the scan.

CMR Postprocessing

All CMR postprocessing was performed blinded to clinical data by an image analyst (K.G.D.) with all analyses verified by an experienced cardiologist (J.H.S.). Ventricular volumes and function were calculated using Medis QMass (MedisSuite 2.1, Medis, Leiden, The Netherlands). Indexed volumes were calculated by dividing by body surface area calculated using the Haycock formula. The presence or absence of LGE, as well as location using the standard 17‐segment model, ²² was qualitatively assessed. Total number of segments with LGE was recorded. Percentage of LGE was calculated using the full‐width half‐maximum (FWHM) technique on the phase‐sensitive inversion recovery images as per our laboratory's standard protocol.

Analysis of myocardial tagged images was performed using harmonic‐phase methodology (Myocardial Solutions, Morrisville, NC) as previously described to calculate circumferential strain (Ecc) at the base, mid, and apex and global Ecc. ²³ T1 maps, obtained before contrast administration, were analyzed with regions of interest manually drawn within the LV myocardium in the standard 16 segments, carefully avoiding partial volume averaging with blood pool or epicardial fat. Areas of LGE were included, as these areas were felt to be the most focal areas in a continuum of diffuse extracellular matrix expansion. ¹⁹ , ²⁴

Risk Score

Our DMD imaging risk score has been described in detail elsewhere. ¹³ Briefly, the score includes the timing and severity of LV dysfunction (<55% and <40%) and the timing and severity of LGE (using the Global Severity Score ¹¹ , ¹⁷ ). Points are awarded on the basis of the age at which a patient develops these findings. Scores cannot decrease but can increase when a patient enters a more severe category at an age young enough to be awarded more points.

Statistical Analysis

Descriptive statistics are presented as median with interquartile range (IQR) for continuous variables and percentages for categorical variables. The modeling strategy was based on those outlined by Harrell. ²⁵ Separate linear models estimated using generalized least squares were used to assess the difference of CMR measurements longitudinally between deceased and alive patients. Each model included fixed effects for future death (versus not), age at CMR measurement modeled flexibly using restricted cubic splines (3 d.f.), and the interaction between age and future death. Correlation among repeated measurements on the same subject were assumed to follow an AR(1) process. AR(1) is an autoregressive process in which the current value is based on the immediately preceding value. We used Wald tests (3 d.f.) to determine if the interaction term of death (versus alive) with age was significant. Residual plots suggested that linear models were appropriate. Descriptive plots of the expected CMR measurements against age by future death were estimated with corresponding 95% CIs to summarize model findings. Age was modeled as a continuous variable to allow for the time period between CMRs to differ between patients, and models used rate of change at the individual level. Individuals with missing CMR measurements, which were minimal, were excluded from analyses for the time points they were missing. The significance level was 0.05.

Results

Study Participants

A total of 63 participants underwent >1 CMR and were included in the analyses (Table 1). The median number of CMR images was 3 (IQR, 2–4; Figure S1), and the median time between CMR imaging was 378 days (IQR, 364–441). At baseline CMR, the median age was 12.5 years (IQR 10.4–15.3; range, 7.4–24.5). There was no significant difference in age at the first CMR between alive and deceased subjects. The majority of patients were already treated with corticosteroids and angiotensin‐converting enzyme inhibitors (Table 1). Twelve subjects (19.1%) died over a median follow‐up of 5 years (IQR, 3.1–7.0). Death occurred at a median age of 17 years. Among deceased patients, 5 (41.7%) had cardiac causes of death, 4 (33.3%) had respiratory causes, 2 (16.7%) had infectious causes, and 1 (8.3%) was unknown. Forty‐four subjects (69.8%) were nonambulatory, and 7 (11.1%) required positive pressure ventilation.

Table 1.

Demographics and Medications at First Visit That Included CMR

	Deceased (n=12)	Alive (n=51)	Overall (N=63)
Age, y	13.2 (10.5–14.0)	11.9 (10.0–15.0)	12.5 (10.1–14.7)
Height, cm	144.4 (123.5–156.2)	145 (125–155)	145 (125–155)
Weight, kg	52.3 (42.8–65.9)	49.1 (33.2–60.8)	49.5 (35.3–61.3)
Body mass index, kg/m2, n (%)
Underweight (<18.5)	1 (8.3)	15 (29.4)	16 (25.4)
Normal weight (18.5–24.9)	5 (41.7)	16 (31.4)	21 (33.3)
Overweight (25.0–29.9)	3 (25.0)	13 (25.5)	16 (25.4)
Obese (≥30)	3 (25.0)	7 (13.7)	10 (15.9)
Current or prior corticosteroid, n (%)			58 (92.1)
Past steroid use	6 (50.0)	13 (25.5)	19 (30.1)
Current steroid use	4 (33.3)	35 (68.6)	39 (61.9)
Current cardiovascular medications, n (%)
Angiotensin‐converting enzyme inhibitor	8 (66.7)	29 (56.9)	37 (58.7)
β Blocker	5 (41.7)	13 (25.5)	18 (28.6)
Angiotensin receptor blocker	2 (16.7)	4 (7.8)	6 (9.5)
Aldosterone inhibitor	1 (8.3)	4 (7.8)	5 (7.9)
Aspirin	1 (8.3)	2 (3.9)	3 (4.8)
Furosemide	1 (8.3)	0 (0)	1 (1.6)

CMR indicates cardiac magnetic resonance. Continuous variables presented as median (IQR).

Baseline Imaging

The median LVEF at the first visit that included CMR was 57%, with 23 subjects (36.5%) having LVEF <55% at this visit (Table 2). Baseline LVEF was higher in alive than deceased subjects (58% versus 54%; P=0.0465). The median baseline LVESVi was 27.0 mL/m² (IQR, 21.2–36.5), and the median LVEDVi was 62.0 mL/m² (IQR, 55.5–76.8). Median Ecc at mid LV was −15.6% (IQR, −17.8 to −13.2), and global Ecc −15.6% (IQR, −17.5 to −13.8). LGE was present in 38 subjects (60.3%), and the median percentage of LGE by FWHM was 26.5% (IQR, 4.88–35.3). Median baseline percentage of LGE by FWHM was not statistically different between alive and deceased subjects (24.1% versus 35.9%; P=0.08). The median baseline CMR risk score was 2 (IQR, 0.4).

Table 2.

Measurements on Initial CMR

	Median (IQR) or n (%)
LVEF, %	57% (IQR, 51.5–60.0)
LVEDV, mL/m2	81 (72 to 104)
LVEDVi, mL/m2	62.0 (55.5 to 76.8)
LVESV, mL/m2	34 (28 to 46)
LVESVi, mL/m2	27.0 (21.2 to 36.5)
Ecc at mid LV (N=57)	−15.6 (−17.8 to −13.2)
Global Ecc (N=56)	−15.6 (−17.5 to −13.8)
LGE present, n (%)	38 (60.3)
Segment of LGE, n (%)
Anterior base	11 (17.5)
Anteroseptal base	6 (9.5)
Inferoseptal base	7 (11.1)
Inferior base	28 (44.4)
Inferolateral base	36 (57.1)
Anterolateral base	34 (54.0)
Anterior mid	13 (20.6)
Anteroseptal mid	6 (9.5)
Inferoseptal mid	12 (19.0)
Inferior mid	30 (47.6)
Inferolateral mid	36 (57.1)
Anterolateral mid	33 (52.4)
Anterior apex	13 (20.6)
Septal apex	17 (27.0)
Inferior apex	17 (27.0)
Lateral apex	21 (33.3)
Apex	11 (17.5)
Number of LGE segments	4.5 (0 to 9.75)
LGE FWHM, % (N=51)	26.5 (4.88 to 35.3)
Native T1 mid LV, ms (N=50)	1051 (1020 to 1085)
CMR risk score	2 (0 to 4)

CMR indicates cardiac magnetic resonance; Ecc, circumferential strain; FWHM, full‐width half‐maximum; LGE, late gadolinium enhancement; LV, left ventricular; LVEDV, left ventricular end‐diastolic volume; LVEDVi, indexed left ventricular end‐diastolic volume; LVEF, left ventricular ejection fraction; LVESV, left ventricular end‐systolic volume; and LVESVi, indexed left ventricular end‐systolic volume.

Longitudinal Analysis

The rate of decline in LVEF was significantly different between alive and deceased subjects (Figure [A]; P<0.001). The rate of increase in LVEDVi (Figure [B]; P=0.0132) and LVESVi (Figure [C]; P<0.001) was also higher in deceased subjects. The rate of increase in LGE, as measured by FWHM, trended toward association with death but did not reach significance (Figure [D]; P=0.17). There was no association between increasing number of LGE segments and death (Figure [E]; P=0.340). Similarly, the rate of change in base T1 (Figure [F]; P=0.22) and mid T1 (Figure [G]; P=0.38) did not differ between groups. Steeper increase in mid Ecc was seen in deceased subjects (Figure [H]; P=0.049), while no difference in global Ecc was seen (Figure [I]; P=0.27). Increase in CMR risk score trended toward association with death but did not reach significance (Figure [J]; P=0.099).

Figure 1. Change in CMR measures in deceased and alive subjects.

Confidence intervals are shown by the shaded gray area. The rate of change in LVEF (A), LVEDVi (B), LVESVi (C), and mid Ecc (H) associates with death while change in LGE FWHM (D), number of LGE segments (E), base T1 (F), mid T1 (G), global Ecc (I), and CMR risk score (J) did not associate with death. * indicates P<0.05. Ecc indicates circumferential strain; FWHM, full‐width half‐maximum; LGE, late gadolinium enhancement; LVEDVi, indexed left ventricular end‐diastolic volume; LVEF, left ventricular ejection fraction; and LVESVi, indexed left ventricular end‐systolic volume.

Discussion

The rate of decline in LVEF and rate of increase in ventricular volumes and mid–LV Ecc associated with all‐cause death in this large cohort of patients with DMD. These CMR measurements could potentially serve as surrogate outcomes for clinical trials, as using death as an outcome measure generally requires prohibitively long follow‐up or initiation of an intervention too late in the disease course. However, additional studies are needed in larger cohorts to validate these findings. Further, identifying patients with rapid rates of change in these metrics may allow for targeted, early intensification of medical therapies in those patients most likely to benefit. Finally, this association with death underlines the importance of serial noninvasive imaging in patients with DMD. Based on our findings, we recommend serial CMR every 1 to 2 years compatible with current DMD care guidelines. ²⁶

LVEF has the most obvious association with death with a clearly faster rate of decline that continues from early in the disease course through death. LVEF measured via CMR has been associated with death in a small number of DMD studies. ⁹ , ¹¹ Our findings expand on this by examining specifically the rate of decline in this measurement over time rather than measurement at a single time point. LVEF has the added benefit of being measurable by echocardiogram when CMR is unavailable. LVEF measured by echocardiogram has been associated with death in some DMD studies. ⁷ , ⁸ , ⁹ , ¹⁰ Importantly, echocardiographic measures of function are less reliable than CMR in DMD, but prior work has demonstrated that echocardiography is adequate for routine clinical evaluation while CMR is best when detection of subtle changes is needed. ²⁷

Rate of change in LVEDVi and LVESVi was similarly higher in deceased subjects compared with living subjects. Two studies have previously demonstrated an association between LVESVi and death, ¹¹ , ¹⁴ and only 1 study has demonstrated an association with LVEDVi. ¹⁴ Both volume metrics increase faster in deceased subjects early in the disease course (second decade) before becoming similar later in the course. This may suggest that these volumetric measurements are an important early measure of cardiac involvement and future death. However, there are also fewer CMR measurements available at later ages, leading to a much larger SD at these time points, which makes interpretation beyond age 20 years less certain. Some studies have observed lower baseline LVEDVi in patients with DMD compared with healthy controls. ²⁸ This underlines the importance of observing the rate of change because 1‐time measures may be falsely reassuring, as they could be in the normal range despite enlargement over time.

Increase in mid–LV Ecc was faster in deceased subjects. LV Ecc has been associated with LV dysfunction ²⁸ , ²⁹ , ³⁰ and death. ¹⁴ Our findings further validate the previous use of Ecc as an outcome in some DMD trials. ³¹ , ³² , ³³ Interestingly, a similar relationship was not observed with global Ecc. These measures seemed to follow overall similar trends but with less separation between deceased and alive subjects with global Ecc. This warrants further investigation in a larger cohort to determine whether this lack of significant difference was related to sample size or represents a true finding.

There was no significant association found with LGE as measured by FWHM or number of segments with LGE. Previous studies have found that severe LGE generally associates with death, ¹¹ , ¹² but this is not a universal finding. ⁹ LGE is considered an early presentation of cardiomyopathy in DMD, often becoming apparent before changes in LVEF. ³⁴ , ³⁵ The median age at first CMR in our cohort was 12.5 years. This raises the possibility that LGE may be worse earlier in the disease course of patients with poor prognosis, such as at the time of first CMR in this cohort, but that rate of change later in the disease course may be less important. If true, this would allow for less contrast use in clinical trials in older patients, a practice that has already been adopted in routine clinical practice in many centers. ³⁶ Additionally, older patients with DMD with advanced disease may have significant discomfort due to contracture or reduced respiratory capacity, which may limit the time available for quality CMR acquisition. Our data indicate that serial noncontrast CMR may be a reasonable alternative to contrast CMR or echo for providing prognostic assessment to patients with DMD, especially once significant LGE is present.

Previous work has shown that our imaging risk score associates with death. ¹³ Despite this association, the present study did not demonstrate an association between change in risk score over time and death. This is likely because the score is designed to be high early in the disease course for those at high risk of early death and then remain largely stable. Similar to our prior study, we found that less than half of patients had a change in score over the study period despite clinical progression of disease.

Limitations

This study is limited primarily by its sample size, and the small number of deaths does not allow for multivariable analyses. The small sample size may also have led to inability to detect some differences, such as the association of change in mid–LV Ecc with death but not global Ecc. Similarly, the small number of deaths does not allow comment on the effect of medications on outcomes. However, this cohort is similar to or larger than other DMD cohorts and was primarily prospective, which provides advantages over other primarily retrospective studies. As some medications, such as angiotensin‐converting enzyme inhibitors, have been shown to change systolic performance in DMD, this may partially confound the results. Additionally, the observational nature of the study does not allow for control of baseline characteristics as seen in the difference in LVEF at the time of first CMR between alive and deceased subjects. We are also limited by the relatively small number of CMR images available in older patients. This leads to substantially decreased power at ages >20 years and may reduce our ability to detect important differences. While the sample size limits the interpretation of the actual rates of change and the range of ages for which these analyses are accurate, the differences between those with and without death are likely accurate and can inform future decisions for clinical trials. We are also limited by the follow‐up time (median, 5 years) in DMD, which is a slowly progressive disease. This prospective observational cohort is ongoing, and future studies will report on longer time periods. Finally, we are not able to provide direct point estimates, such as odds ratios or hazard ratios, for slope values. This is because any such point estimates would require a somewhat arbitrary assignment of an intercept value, making interpretation difficult. Additionally, relationships between age and CMR measurements were not linear, making calculation of a single slope infeasible.

Conclusions

DMD all‐cause death is associated with the rate of change of LVEF, mid Ecc, and ventricular volumes. Aggressive medical therapy to decrease the rate of progression may improve the mortality rate in this patient population. In addition, a decrease in the rate of progression may serve as a valid surrogate outcome measure for therapeutic trials.

Sources of Funding

Research reported in this publication was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under Award Number K23HL123938 and R56HL141248 (Bethesda, MD) (Dr Soslow). The project was supported by the National Center for Research Resources, Grant UL1 RR024975‐01, and is now at the National Center for Advancing Translational Sciences, Grant 2 UL1 TR000445‐06 (Bethesda, MD). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This project was supported by the Food and Drug Administration Grant R01FD006649 (Dr Soslow). This project was supported by the Fighting Duchenne Foundation and the Fight DMD/Jonah & Emory Discovery Grant (Nashville, TN) (Dr Soslow) and the Ackerman/Nicholoff Family Foundation (Indianapolis, IN) (Dr Markham). Dr Starnes's project was supported by grant number T32 HS026122 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. The sponsors and funders had no role in the design and conduct of the study or in the collection, analysis, and interpretation of the data or in the preparation, review, or approval of the manuscript. There are no relationships with industry or other entities to declare.

Disclosures

None.

Supporting information

Figure S1