Exploring the natural history of bone mineral density in Duchenne muscular dystrophy: a systematic literature review

Abstract

Purpose

In Duchenne muscular dystrophy (DMD) osteoporosis and fragility fractures present significant comorbidities, resulting from progressive myopathy reduced weight-bearing and osteotoxic effects of prolonged glucocorticoid (GC) use. This systematic literature review (SLR) assessed the natural course of bone mineral density (BMD) in individuals with DMD and the factors associated with its natural progression.

Methods

A systematic literature review was conducted to evaluate BMD in individuals with DMD. Searches were performed in PubMed® and Embase® for peer-reviewed articles published between 2000 and 2023, focusing on DMD patients aged 5-15 years receiving GC treatment. Statistical analyses employed an initial correlation analyses followed by generalized linear mixed models (GLMMs) to investigate relationships between bone health biomarkers and clinical factors.

Results

Strong associations were observed between lumbar spine (LS) BMD Z-scores, age and GC treatment duration. LS areal BMD (aBMD) Z-scores declined with age (–0.25 SDS/year, p < 0.001, n = 1527) and with longer GC use (–0.24 SDS/year, p < 0.001, n = 817). Total body less head (TBLH) and lateral distal femur (LDF) BMD Z-scores also decreased with age at a rate of –0.26 SDS/year and – 0.41 SDS/year, respectively. The percentage of individuals with fractures increased by 21% (p < 0.01, n = 1066) for every 1-SDS decrease in LS aBMD Z-score. Reductions in LS aBMD Z-scores significantly accelerated after 11 years of age, with older versus younger individuals experiencing a faster decline.

Conclusion

Understanding the natural history of BMD and disease progression relating to bone fragility will help to support clinical trial design to assess and improve bone strength in DMD. To our knowledge, this is the first SLR investigating BMD Z-score trajectories in individuals with DMD receiving GCs and highlights associations between age and GC treatment duration with declining LS aBMD, TBLH, and LDF Z-scores.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00198-025-07651-6.

Introduction

Duchenne muscular dystrophy (DMD) is an X-linked genetic disorder caused by pathogenic variants in the DMD gene, resulting in the absence of functional dystrophin protein and progressive muscle degeneration [1]. The current standard of care in DMD includes glucocorticoids (GCs) which improve muscle strength and mobility, and delay loss of ambulation (LoA) [2–4]. DMD primarily affects muscle tissue, leading to muscle wasting, but also has notable effects on bone due to the close relationship between muscle and bone strength (the “muscle–bone unit”) [5]. In addition, the prolonged use of GCs has a strong osteotoxic effect further exacerbating bone impairments observed in DMD [6, 7].

Bone morbidity, including osteoporosis and fragility fractures, is a major complication in individuals with DMD [8]. Boys with DMD frequently develop osteoporosis and low bone mineral density (BMD) due to the progressive myopathy, reduced weight-bearing, and the prolonged use of GCs [6, 7]. Individuals with DMD are highly susceptible to low-trauma (fragility) fractures, even with BMD Z-scores in the normal range, but fractures markedly increase as BMD decreases [9]. While long bone fractures are the most frequently reported fractures in the DMD population, vertebral fractures (VFs) show a far greater relative increase in incidence compared with healthy children. Furthermore, VFs are often asymptomatic, resulting in under-reporting in clinical studies. A study by Bothwell et al. showed that almost all individuals with DMD will develop VFs if they remain on daily GC regimens [10]. The risk of long bone fractures in individuals with DMD is estimated to be at least 4 times higher than in healthy children [11], but for VFs this is more than 500 times higher [11], since VFs are rare in the healthy pediatric population. Fracture incidence in DMD is normally age dependent, peaking between 14 and 15 years of age [11].

This SLR investigated the natural course of BMD in individuals with DMD with mean ages between 5 and 15 years receiving GC treatment, and explored how BMD trajectories are influenced by various patient characteristics. The outputs from this SLR intend to improve the understanding of osteoporosis progression and provide insights into the timeline of bone health events, in individuals with DMD. This information will support the design of future clinical trials assessing bone strength in individuals with DMD by helping to identify relevant endpoints such as the rate of decline of total body less head (TBLH), lateral distal femur (LDF), lumbar spine areal BMD (LS aBMD) as well as bone mineral apparent density (BMAD) Z-scores, and risk factors for fractures. It is important to note that aBMD is an areal bone density (i.e., the ratio of bone mineral content to projected area with units of g/cm²), BMAD represents a size-corrected BMD that approximates vBMD in g/cm³ units, whereas true vBMD represents a measured volumetric bone density, also with units of g/cm³ and captured by three-dimensional quantitative computed tomography (QCT) [12–14]. One of the important distinctions between BMAD and vBMD is that BMAD includes both cortical and trabecular bone (including the posterior spinal processes) like aBMD, whereas vBMD is only measuring trabecular bone [15]. These three measures are described and analyzed separately in this SLR.

Methods

A comprehensive SLR was conducted to evaluate the available data on BMD in individuals with DMD using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) as a guide [16].

Search strategy

Scientific articles published in peer-reviewed journals between January 1st, 2000 and December 31st, 2023 assessing BMD in individuals with DMD were identified through searches in PubMed® and Embase®. The search strategy included combinations of terms related to “bone density”, “bone mineral density”, “Duchenne muscular dystrophy”, or “DMD”. The search strategies used are detailed in Table SI1. The first record describing BMD in DMD was published in 2000 by Larson CM et al. [17], setting the earliest year for our search strategy.

Selection criteria

Eligibility criteria were based on the Population, Interventions, Comparators, Outcomes, and Study (PICOS) design framework [18]. The inclusion and exclusion criteria specified English-language publications and were designed to include articles reporting studies assessing individuals with DMD only, with mean ages between 5 and 15 years, who were receiving GC treatment in the absence of bisphosphonate or testosterone treatment. Outcomes of interest included aBMD (which has been standardized by age-and sex) Z-scores of the LS, bone size corrected LS aBMD (i.e., BMAD), and vBMD acquired by QCT, TBLH, and LDF or proximal femur (PF, also known as total hip). Only studies reporting original data were included. Full inclusion and exclusion criteria for article type and patients included in the SLR are shown in Table 1. Bone size correction for BMD refers to the process of correcting aBMD measurements to account for variations in bone size using the Carter [13] or Kroger [19] methods described elsewhere. This approach is particularly important in a cohort with the potential for short stature (since aBMD by DXA is under-estimated in those with short stature). By correcting for bone size with respect to DXA-based aBMD measurements, this strategy will provide a more precise and accurate assessment of BMD that, in turn, will better predict bone strength and fracture risk [13].

Table 1.

Inclusion and exclusion criteria for the SLR

Criteria	Details
	Inclusion	Exclusion
Population	• Individuals diagnosed with DMD • Mean age at baseline ≥ 5 to < 16 • All or a vast majority of individuals receiving steroids, where available only those receiving daily steroids • Individuals with aBMD, BMAD, and/or vBMD Z-scores of the LS, TBLH, LDF, and/or PF/total hip BMD Z-scores • Also included: individuals receiving placebo in clinical trials assessing bisphosphonate or anabolic treatments and baseline values before treatment	• Individuals treated with bisphosphonates and/or anabolic drugs (e.g., testosterone, teriparatide) • Individuals not on GCs (explicit in the paper) • Individuals treated with intermittent GCs (explicit in the paper or where differentiated vs daily) • Other diseases and neuromuscular disorders (e.g., SMA, Becker muscular dystrophy, cerebral palsy)
Intervention	• Steroids as a standard of care for DMD, where available only daily steroids
Comparator	• N/A
Outcomes	• aBMD Z-scores of the LS • BMAD Z-scores of the LS • vBMD Z-scores of the LS acquired by QCT • aBMD Z-scores of the TBLH • aBMD Z-scores of the LDF and/or PF/total hip
Study designs	Inclusion	Exclusion
	• Randomized controlled trials • Prospective studies • Cross-sectional studies • Retrospective longitudinal studies • Retrospective chart reviews • Case seriesa	• Reviews • Case studies • Non-clinical studies • Congress abstracts • Book chapters
Language	English
Country	Any
Date of publication	Published (available online) between January 2000 and December 2023

aBMD areal BMD, BMAD bone mineral apparent density, BMD bone mineral density, DMD Duchenne muscular dystrophy, GC glucocorticoid, LDF lateral distal femur, LS lumbar spine, N/A not applicable, PF proximal femur, QCT quantitative computed tomography, SLR systematic literature review, SMA spinal muscular atrophy, TBLH total body less head, vBMD volumetric BMD

^aWith five or more individuals

Inclusion and exclusion criteria were defined before the review process.

Screening, data extraction, and synthesis

Two independent reviewers screened and identified articles based on the predefined eligibility criteria (CDF, MG). Any disagreement between the reviewers was resolved by consensus. Records were first screened based on title and abstract to eliminate duplicates, then subsequently screened at a full-text level. Data were extracted by one reviewer (CDF) and validated by a second independent reviewer (MG). The variables to include in the analysis and the meaningfulness of the results were reviewed and discussed with all the authors of this SLR (CDF, MG, YC, APM, CW, HJM, EM, NC, LMW).

Baseline characteristics were extracted from the articles, and when available, longitudinal data and sub-group analyses were analyzed. BMD data as described in the Outcomes section of Table 1 were extracted and assessed. In the records included in this SLR, long bone fractures were described according to their anatomical site and frequency and VFs were described according to their location, type, and severity (adjudicated according to a standard method such as Genant’s semiquantitative method). All fracture information is included in the supplementary information.

Statistical analysis

All statistical analyses were conducted using R (version 4.4.1). Data was taken as means from each study or sub-population of a study and plotted as individual points for the regression analyses. Study-level summary statistics were used to investigate relationships between bone health biomarkers and relevant clinical and demographic factors. The primary bone biomarkers analyzed included LS aBMD, BMAD, LS vBMD, TBLH BMD, and femur BMD Z-scores (proximal and lateral distal). Clinical manifestations of interest included ambulation status and incidence of fractures, while key covariates included mean age and duration of GC treatment.

A two-stage analytical approach was employed. First, correlation analyses were conducted using both Pearson and Spearman methods to assess the strength and direction of relationships between variables and to evaluate robustness against potential outliers. These exploratory results informed the development of refined models in the second stage. Generalized linear mixed models (GLMMs) were then fitted using the lme4 package, incorporating a random intercept for study to account for intra-study clustering. Association analyses were conducted in three tiers: (1) Between bone biomarkers and covariates of interest, (2) Between bone biomarkers and clinical manifestations, and (3) Between clinical manifestations and covariates.

Weights were primarily based on inverse-variance (derived from reported standard deviations [SD] and sample sizes); if this approach reduced data coverage, models were re-estimated using sample size as weights. Model selection between GLMMs and standard linear regression was guided by diagnostics, including intraclass correlation coefficients (ICC) and singular fit checks. Subgroup and interaction models were further applied to explore heterogeneity, particularly across age categories and fracture status. All statistical modeling and visualizations, including fitted regression lines and annotated model equations, were generated using ggplot2.

All authors contributed to the development and review of this SLR. YC performed the statistical analyses.

Results

Literature review

A total of 525 published articles were retrieved from the Embase® search and 152 from the PubMed® search using the specified search terms. After the removal of duplicates and congress abstracts, 241 records were screened for inclusion in the analysis (Fig. 1). The references of the included records were reviewed to find potential missed records. Two more records were added manually [20, 21].

Fig. 1.

PRISMA flow diagram showing records included and excluded from the final analyses. PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses

Following the screening process plus manual additions, a total of 41 articles met the criteria for inclusion. Of the 41 articles included, 16 reported on prospective studies including two randomized controlled trials, 17 reported on retrospective studies, and eight on cross-sectional studies (Table 2). Eight studies included longitudinal data with more than one dual-energy x-ray absorptiometry (DXA) measurement per individual [22–29]. Only one record reporting QCT-acquired vBMD included longitudinal data [30].

Table 2.

Key characteristics of studies included in the analyses

Author and year [country] [Reference]	Study design	Sample, n	Age mean (SD), y	BMD Z-scores
Al-Zougbi et al. 2019 [USA] [31]	Retrospective chart review	5	13.7y, range 11.9–14.1y	LS (areal): −2.6
Annexstad et al. 2019 [Norway] [32]	Prospective longitudinal observational study	18	9y, range 6–12y	LS (areal): −1.19; TBLH: −2.02
Annexstad et al. 2019 [Norway] [32]	Prospective longitudinal observational study	12	15y, range 13–18y	LS (areal): −2.94; TBLH: −3.03
Barzegar et al. 2018 [Iran] [33]	Cross-sectional	All patients: 30	11 y	LS (areal): −1.5; PF: −3.4
Barzegar et al. 2018 [Iran] [33]	Cross-sectional	Ambulatory: 21		LS (areal): −1.5; PF: −3.4
Barzegar et al. 2018 [Iran] [33]	Cross-sectional	Non-ambulatory: 9		LS (areal): −1.4; PF: −3.4
Bianchi et al. 2011 [Italy] [34]	Prospective longitudinal	BL: 33	8.4y, range 5–13y	BL BMAD: −2.3 12-month BMAD: −2.8 BL TBLH: −2.1 12-month TBLH: −2.6
Catalano et al. 2022 [Italy] [35]	Cross-sectional observational study	All patients: 31	14y, range 12–21.5y	LS (areal): −3.69
Catalano et al. 2022 [Italy] [35]	Cross-sectional observational study	Ambulatory: 11	8y, range 7–12y	LS (areal): −1.5
Crabtree et al. 2010 [UK] [36]	Prospective longitudinal interventional	BL: 25	8.9y, range 5.5–12y	TBLH: −3 LS (areal): −1.4 BMAD: −0.9
Crabtree et al. 2018 [UK] [37]	Retrospective pragmatic longitudinal study	Daily steroid regime	BL: 19 (8.1y)	TBLH: −2.3; BMAD: −0.6
Crabtree et al. 2018 [UK] [37]	Retrospective pragmatic longitudinal study	12 months: 19		TBLH: −2.4 BMAD: −0.6
Crabtree et al. 2018 [UK] [37]	Retrospective pragmatic longitudinal study	24 months: 19		BLH: −3.1 BMAD: −0.4
Doulgeraki et al. 2016 [Greece] [38]	Cross-sectional	42	9.5y	TBLH: −1.8 LS (areal): −1.2
Escolar et al. 2011 [USA] [24]	RCT	Daily steroid regime BL: 32	7.5y, range 4–10y	LS (areal): −1.1
Escolar et al. 2011 [USA] [24]	RCT	Daily steroid regime: 12 months		LS (areal): −1.3
Harcke et al. 2006 [USA] [9]	Retrospective	Ambulatory: 41	8.9y	LS (areal): −1.8; LDF: −1.5
Harcke et al. 2006 [USA] [9]	Retrospective	Partially ambulatory: 16	11.8y	LS (areal): −2.1; LDF: −2.6
Harcke et al. 2006 [USA] [9]	Retrospective	Non-ambulatory: 25	12.6y	LS (areal): −2.0; LDF: −3.6
Hawker et al. 2005 [Canada] [39]	Prospective before –after trial	BL: 42	10.8y, range 6.9–15.6y	LS (areal): −1.94
Henderson et al. 2010 [USA] [21]	Retrospective	112	11.8y, range 6–18y	LS (areal): −1.7 LDF: −3.8
Houde et al. 2008 [Canada] [25]	Longitudinal, retrospective chart review	BL: 37	(7.6y)	LS (areal): −1.8
Houde et al. 2008 [Canada] [25]	Longitudinal, retrospective chart review	7-year FU: 38	13.3y	LS (areal): −4.5
Houston et al. 2014 [USA] [40]	Retrospective cohort study	BL: 29	11.1y	LS (areal): −2.04 Total hip: –3.30
Kervin et al. 2018 [USA] [41]	Retrospective chart review	32	11.3y	LS (areal): −2.2
King et al. 2014 [USA] [42]	Prospective Descriptive	22	11.5y, range 5–17y	TBLH: −2.3
Lee et al. 2020 [Australia] [43]	Retrospective, observational, pre–post study	BL before testosterone: 16	14.5y, range 14–17.7y	LS (areal): −3.29
Liaw et al. 2023 [Australia] [20]	Retrospective cohort study Chart review	All patients: 155		NR
Liaw et al. 2023 [Australia] [20]	Retrospective cohort study Chart review	With fractures: 71	11.4y, IQR 5.3	LS (areal): −2.4
Lim et al. 2017 [Australia] [44]	Retrospective cohort study Chart review	BL before BPs: 9	12.6y, range 9–14.6y	LS (areal): −3.03 BMAD: −0.94
Liu et al. 2022 [China] [30]	Retrospective longitudinal study	BL: 19	8.58y	LS (vol): −0.85
Liu et al. 2022 [China] [30]	Retrospective longitudinal study	12 months: 19	9.69y	LS (vol): −1.56
Liu et al. 2022 [China] [30]	Retrospective longitudinal study	24 months: 19	11.15y	LS (vol): −2.02
Liu et al. 2022 [China] [30]	Retrospective longitudinal study	36 months: 3	12.73y	LS (vol): −2.41
Ma et al. 2017 [Canada] [26]	Retrospective longitudinal study	GC initiation: 30	7.8y	LS (areal): −1.2; LS (vol): −0.4
Ma et al. 2017 [Canada] [26]	Retrospective longitudinal study	12 months: 30		LS (areal): −1.1; LS (vol): −0.3
Ma et al. 2017 [Canada] [26]	Retrospective longitudinal study	FU: 30	6.9y	NR
Mayo et al. 2012 [Canada] [27]	Prospective longitudinal study	BL: 39	6.6y	LS (areal): −1.1
Mayo et al. 2012 [Canada] [27]	Prospective longitudinal study	1–2y: 36	8.0y	LS (areal): −1.5
Mayo et al. 2012 [Canada] [27]	Prospective longitudinal study	3–4y: 24	10.3y	LS (areal): −1.8
Mayo et al. 2012 [Canada] [27]	Prospective longitudinal study	5–6y: 25	11.8y	LS (areal): −2.4
Mayo et al. 2012 [Canada] [27]	Prospective longitudinal study	7–8y: 13	14.2y	LS (areal): −3.6
McAdam et al. 2012 [Canada] [45]	Retrospective cohort study Chart review	Montreal BL: 37	7.6y	BL: NR; 12 months, LS (areal): −1.8
McAdam et al. 2012 [Canada] [45]	Retrospective cohort study Chart review	Toronto BL: 40	7.7y	LS (areal): −1.1
Misof et al. 2016 [Austria] [46]	Prospective before − after trial	BL before BPs: 9	11.4y	LS (areal): −1.9 LS (vol): −0.7
Naume et al. 2023 [Denmark] [47]	Prospective observational study	23	9.3y, range 4–16	LS (areal): −1.7
Phung et al. 2023 [Canada] [48]	Cross-sectional	60	11.5y, range 5.4–19.5	TBLH: −2.4 LS (areal): −2.4 LS (BMAD): −2.1 LDF: −4.0
Phung et al. 2024a [Canada] [22]	Prospective study	BL: 38	11.0y	TBLH: −1.9; LS (areal): −2.1; LS (vol): −1.9; LDF: −3.7
Phung et al. 2024a [Canada] [22]	Prospective study	12 months: 38		TBLH: −2.3; LS (areal): −2.3; LS (vol): −2.1; LDF: −3.7
Sbrocchi et al. 2012 [Canada] [49]	Retrospective observational study	BL before BPs: 7	11.6y, range 8.5–14.3y	LS (areal): −2.1; LS (vol): NR
Sertpoyraz et al. 2019 [Turkey] [50]	Cross-sectional descriptive study	55	13.9y, range 8–25y	LS (areal): −2.0 Total hip: –3.2
Singh et al. 2018b [Canada] [51]	Retrospective chart review	49	14.2y at final assessment	Initial LS (areal): −1.5 Initial TBLH: −1.8 Final LS (areal): −1.9 Final TBLH: −2.5
Söderpalm et al. 2007 [Sweden] [52]	Cross-sectional study	24	11.9y, range 2.3–19.7y	LS (areal): −2.5
Söderpalm et al. 2012 [Sweden] [28]	Prospective observational longitudinal study	< 10y BL: 8	Range 2.3–19.7y (6.6y)	LS (areal): −0.7
Söderpalm et al. 2012 [Sweden] [28]	Prospective observational longitudinal study	< 10y 4y FU: 8	10.4y	LS (areal): −1.2
Söderpalm et al. 2012 [Sweden] [28]	Prospective observational longitudinal study	> 10y BL: 10	13.9y	LS (areal): −3.0
Söderpalm et al. 2013 [Sweden] [29]	Prospective longitudinal study	BL before WBV: 6	6.8y, range 5.7–12.5y	LS (areal): −0.9
Suthar et al. 2021 [India] [53]	Prospective observational study	76 54 underwent DXA scans	8.5y, IQR 7.04–10.77	LS (areal): −2.3 PF: −2.5
Thandrayen et al. 2021 [South Africa] [54]	Prospective. Descriptive	10	13.5y	TBLH: −3.8
Tian et al. 2016 [USA] [55]	Retrospective chart review	FMS 1: 134	7.6y, range 5.0–14.1y	LS (areal): −1.63, LDF: −2.45
Tian et al. 2016 [USA] [55]	Retrospective chart review	FMS 2: 163	8.6y, range 5.0–16.1y	LS (areal): −1.41, LDF: −2.77
Tian et al. 2016 [USA] [55]	Retrospective chart review	FMS 3: 84	10.8y, range 5.5–18.0y	LS (areal): −1.39, LDF: −3.67
Tian et al. 2020 [USA] [56]	Retrospective chart review	BL before BPs: 54	11y, range 5.5–16.6y	LS (areal): −2.13 (n = 51) LDF: −2.70
Tsaknakis et al. 2022 [Germany] [57]	Cross-sectional cohort study	37	15.6y	LS (vol): −4.1
Tung et al. 2021 [Hong Kong] [58]	Cross-sectional, case–control study	9	9.3y, IQR 8.5–10.8	LS (areal): −1.3 TBLH: −4.4
Zacharin et al. 2021 [Australia] [23]	RCT	BL pbo: 31	10.1y, range 6–16y	LS (areal): −1.8
Zacharin et al. 2021 [Australia] [23]	RCT	12-month pbo: 31		LS (areal): −2.1
Zacharin et al. 2021 [Australia] [23]	RCT	24-month pbo: 31		LS (areal): −2.1
Zacharin et al. 2021 [Australia] [23]	RCT	BL before ZA: 31	10.1y, range: 6–16y	LS (areal): −2.0
Zheng et al. 2020 [China] [59]	Prospective longitudinal	BL before ZA: 17	10.9y	LS (areal): + 0.1, PF: −2.7
Zheng et al. 2020 [China] [59]	Prospective longitudinal	BL before ALN: 18	9.7y	LS (areal): + 0.1, PF: −2.3
Zheng et al. 2020 [China] [59]	Prospective longitudinal	BL control: 17	8.2y	LS (areal): + 0.6, PF: −1.0

ALN alendronate, BL baseline, BMAD bone mineral apparent density, BMD bone mineral density, BPs bisphosphonates, FMS functional mobility score, FU follow-up, GC glucocorticoid, IQR interquartile range, NR not reported, LDF lateral distal femur, LS lumbar spine, Pbo placebo, PF proximal femur, RCT randomized controlled trial, SLR systematic literature review, TBLH total body less head, Vol volumetric, WBV whole body vibration, ZA zoledronic acid, y years

^aThis record was made available online before the December 31^st 2023, meeting the inclusion criteria for this SLR

^bPamidronate was prescribed to 73% of those who sustained VF at some point with a BMD Z-score of < − 2.4; however, no uniformity could be seen as to the time or event when pamidronate was started in these patients; for more details, refer to the record

^cThis is the only study using specific Chinese reference data, which differs from all the other studies using more broadly used reference data; the LS BMD Z-score data were not used for the main analysis

Overall, data from 1940 individuals were included in the analysis. However, not all articles reported the same variables; therefore, the number of observations (N) per analysis or characteristic varied and was specifically reported. The included articles covered data collected across studies that took place in five continents: Africa (n = 1), Asia (n = 6), Australia (n = 4), Europe (n = 12), and North America (n = 18), minimizing the location bias (Table 2). Additional details of the 41 articles included are presented in Table SI2.

Regarding BMD, 39/41 records described LS BMD, 21 records described only LS, seven records described LS and TBLH, nine records described LS and femur with four LDF and five PF/total hip, and two records described LS, TBLH, and LDF. Only two records described TBLH BMD alone and no records described either LDF or PF BMD alone (Fig. SI1). The most common scanner platform used for obtaining BMD in individuals with DMD (according to this SLR) was the DXA scanner (39/41); in the remaining two studies a QCT scanner was used [30, 57]. For the 39 records using DXA scanners, the most commonly used was the Hologic in 16/39 records (as the only available scanner), the GE Lunar was used in 11/39 records (as the only available scanner), both platforms were used in 4/39 records, and 8/39 records did not describe the equipment used. Table 2 and SI2 summarize all the main characteristics of the included records.

Of the 39 records describing LS BMD Z-score, most of them described only aBMD (29/39), two records reported correcting for bone size, without mentioning LS aBMD Z-score [34, 37], two records acquired true vBMD by QCT [30, 57], and six records reported both aBMD and BMAD [26, 36, 44, 46, 48, 60], for a total of eight records reporting BMAD, and two records describing vBMD. There were two methods for bone size correction of LS aBMD Z-score (Carter method [13], Kroger method [61]). From the eight records reporting BMAD, six used the Carter method, two used the Kroger method, and two did not specify the method used. Table SI3 shows the methods of correction and equipment used by each of the records reporting BMAD.

Baseline characteristics

The baseline characteristics for the 1940 individuals with DMD across the 41 studies are shown in Table 3. The sample sizes ranged from 5 to 381 individuals, justifying weighing of the analyses by sample size. The median age at baseline was 10.10 years, ranging from 6.6 to 15.6 years. Of the eight and ten studies that reported pubertal and ambulatory status, respectively, most were prepubertal (71%) and ambulatory (72%). Details of GC treatment duration were reported in 24 studies, covering 1213 individuals, with a median treatment duration of 35.4 months, ranging from 0 (start of GC treatment) to 8.76 years. The median dose of daily GCs (prednisone equivalent) was 0.66 mg/kg/day (range 0.22–0.75) reported in 25 studies covering 931 individuals, representing 88% of the recommended dose (0.75 mg/kg/day). The bone age delay was reported in seven instances, showing a median delay of 1.20 years (range 0.20–1.70). The median body mass index was 19.60 kg/m² (range 16.10–23.20; 13 studies), which is higher than the 50th centile in children aged 10 years (16.50 kg/m²) according to World Health Organization reference data [62]. Both factors are closely related to the long-term use of GCs.

Table 3.

Baseline characteristics of individuals included in the analyses

Baseline characteristic	Mean (SD)	Median (min, max)	N	Studies
Age (years)	10.14 (2.82)	10.10 (6.60, 15.60)	1940	41
Age at diagnosis (years)	5.06 (2.39)	4.45 (3.00, 9.30)	585	12
Age at start of GC (years)	6.60 (1.88)	6.80 (5.00, 8.90)	850	19
Age at LoA (years)	10.96 (2.18)	10.75 (9.10, 11.70)	415	10
% non-ambulatory	35.95 (25.50)	28.00 (0.00, 100.00)	452	10
Duration of GC (months)	33.74 (12.97)	35.40 (0.00, 105.12)	1213	24
GC dose (mg/kg/day)a	0.58 (0.14)	0.66 (0.22, 0.75)	931	25
BMI (kg/m2)	19.57 (3.49)	19.60 (16.10, 26.40)	376	13
Bone-age delay (years)	1.15 (1.90)	1.20 (0.20, 1.70)	218	7
Tanner stage 1 (%)	74.05 (24.59)	70.97 (0.00, 100.00)	297	8
Tanner stages 2–5 (%)	25.95 (24.54)	29.03 (0.00, 100.00)	297	8
% individuals with history of fractures	35.60 (21.10)	29.41 (0.00, 97.00)	1546	26
LS aBMD Z-score	− 1.88 (0.90)	− 1.85 (− 3.69, − 0.70)	1527	34
BMAD Z-score	− 0.99 (0.70)	− 0.7 (− 2.10, − 0.30)	223	8
vBMD Z-score	− 2.20 (1.24)	− 2.02 (− 4.17, − 0.85)	56	2
TBLH BMD Z-score	− 2.38 (1.45)	− 2.40 (− 4.40, − 1.80)	399	11
LDF BMD Z-score	− 3.10 (1.24)	− 3.60 (− 4.00, − 1.50)	727	6
PF/total hip BMD Z-score	− 2.63 (0.83)	− 2.70 (− 3.4, − 1.00)	220	5

aBMD areal BMD, BMAD bone mineral apparent density, BMD bone mineral density, BMI body mass index, GC glucocorticoid, LoA loss of ambulation, LDF lateral distal femur, LS lumbar spine, min minimum, max maximum, PF proximal femur, QCT quantitative computed tomography, SD standard deviation, TBLH total body less head, vBMD volumetric BMD, VF vertebral fracture

^aPrednisone equivalent

Baseline BMD Z-scores at different anatomical sites showed a considerable reduction compared with matched controls (Table 3). LDF BMD Z-score appeared to be the most affected site (below − 3 standard deviation score [SDS]), followed by PF/total hip, TBLH (both below − 2 SDS), and then LS (aBMD, BMAD, and vBMD Z-scores). Of note, the percentage of individuals with a history of fractures was reported in 26 studies, including 1546 individuals, with a median of 29.41% (range 0–97%; Table 3).

Association of LS BMD Z-scores with age in individuals with DMD

LS aBMD Z-scores were reported across 34 studies (Table 3). A statistically significant linear correlation between LS aBMD Z-score and age was observed (Pearson’s correlation coefficient: − 0.8658, p < 0.001). After refinement of the results using the GLMM approach, the LS aBMD Z-score was shown to decline with age at a rate of − 0.25 SDS per year (95% confidence interval [CI] [− 0.30, − 0.20]; p < 0.001, n = 1527, 34 studies; Fig. 2a, b). Most data obtained were from studies of individuals with DMD who had mean ages between 8 and 11 years, with a consistent decline in the LS aBMD Z-score observed as age increased.

Fig. 2.

Association between BMD Z-scores at different anatomical sites with age and GC treatment duration in individuals with DMD. a Mean LS BMD Z-scores in individuals with DMD plotted against mean age. b Mean LS BMD Z-scores for different age groups. c LS BMD Z-score trajectories in individuals with DMD 5–10 years of age versus 11–15 years of age. d Mean LS BMD Z-scores plotted against mean duration of GC duration. e Mean TBLH BMD Z-score versus mean age in individuals with DMD. f Mean TBLH and femur (LDF and PF) BMD Z-scores for different age groups. Datapoints represent either a study or a population, sub-group, or longitudinal data from a study. For Fig. 2a, d blue dots represent LS aBMD and red dots represent BMAD. Blue lines represent the regression model for aBMD. The red lines represent the regression model for BMAD. aBMD areal BMD, BMAD bone mineral apparent density, BMD bone mineral density, DMD Duchenne muscular dystrophy, GC glucocorticoid, GLMM generalized linear mixed model, LDF lateral distal femur, LS lumbar spine, NS not significant, PF proximal femur, SDS standard deviation score; TBLH total body less head, vBMD volumetric BMD

The mean LS aBMD Z-score fell below − 2 for individuals with DMD by a mean age of 10.6 years, according to the linear regression, suggesting worsening LS aBMD Z-score at a young age in individuals with DMD (Fig. 2a, b). These results are aligned with the literature where the mean age for an LS aBMD Z-score of ≤ − 2.0 was reported as 10.5 years [63]. BMAD also showed a statistically significant linear correlation with age (Pearson’s correlation coefficient: − 0.6467, p < 0.0352). Using the GLMM approach, the association with age was not statistically significant (p = 0.7158, n = 223, 8 studies; Fig. 2a, b).

Additionally, the LS aBMD Z-score significantly decreased during the first year after starting GC treatment and continued to decrease at a slower rate until 11 years of age (Figs. 2c and SI2). Age greater than 11 years was significantly associated with an accelerated rate of decline in LS aBMD Z-scores compared with a younger age (of 5–10 years, p < 0.001; Fig. 2c).

The mean decrease in LS aBMD Z-score changed from − 0.10 SDS per year (95% CI [− 0.18, − 0.02]; p < 0.01) in younger individuals (5–10 years of age) to − 0.34 SDS per year in individuals aged 11–15 years (95% CI [− 0.47, − 0.21]; p < 0.001; Fig. 2c). The difference between the two groups was highly significant (p < 0.001).

Insufficient data on BMAD and vBMD were available to calculate accurate rates of decline with age. Nevertheless, the available data suggest that BMAD and LS vBMD decline with age (Fig. 2b).

Association of LS BMD Z-scores with GC treatment duration

Duration of GC treatment was reported in 15 LS aBMD studies, six BMAD studies, and one LS vBMD study. Across the studies, GC treatment was started at a median (min, max) age of 6.80 (5.00, 8.90) years (Table 3). Statistically significant correlations were observed for LS aBMD (Pearson’s correlation coefficient: − 0.7868, p < 0.001). A strong association was also observed using the GLMM approach; for every 1-year increase in the duration of GC treatment, the LS aBMD Z-score decreased by 0.24 SDS (95% CI [− 0.16, − 0.32]; p < 0.001, n = 817, 15 studies; Fig. 2d). In the case of BMAD, the linear correlation did not reach statistical significance (Pearson’s correlation coefficient: − 0.5451, p = 0.1032). The GLMM approach showed a similar trend (p = 0.6991, n = 181, six studies). Insufficient data on BMAD and vBMD were available to calculate accurate rates of decline with GC treatment duration.

Association of TBLH and femur (LDF and PF) BMD Z-scores with age in individuals with DMD

Published data on TBLH and femur (LDF and PF) BMD Z-scores in individuals with DMD were less frequently reported compared with LS BMD Z-scores. TBLH BMD Z-score was reported by 11 studies, LDF BMD Z-score by six studies, and PF/total hip by five studies (Table 3). TBLH Z-scores (Fig. 2e, f) showed a statistically significant linear correlation with age (Pearson’s correlation coefficients: − 0.6044 [p < 0.01]). After refining the analysis with the GLMM approach, a statistically significant association with age was observed. For every 1-year increase in mean age, TBLH BMD Z-score decreased by 0.26 SDS (95% CI [− 0.35, − 0.17]; p < 0.001, n = 399, 11 studies). LDF Z-scores showed a statistically significant linear correlation with age (Pearson’s correlation coefficients: − 0.6674 [p = 0.0248]). The GLMM approach showed a statistically significant association with age. For every 1-year increase in mean age, LDF BMD Z-score decreased by 0.41 SDS (95% CI [− 0.57, − 0.26]; p < 0.001, n = 727, 6 studies), being the anatomical site with the fastest rate of decline in DMD (Fig. 2g).

Insufficient data were obtained for PF BMD Z-score to accurately calculate a rate of decline; however, the means shown in Table 3 and Fig. 2f suggest that the PF BMD Z-score is also considerably compromised in individuals with DMD.

Association of BMD Z-scores in individuals with fractures and without fractures

Out of the 41 records included in this SLR, nine reported differences in aBMD Z-scores between individuals with fractures and without fractures (Table SI4). The mean age of the individuals as reported on the records was correlated with their corresponding mean aBMD Z-scores, which was not necessarily only longitudinal data, but also cross-sectional. Individuals with fractures (fractures occurring within the reported study timeframe or described at a timepoint on cross-sectional studies) showed a progressive and steep decline in the lumbar spine areal BMD (LS aBMD) Z-score as age increased compared with those individuals without fractures. A strong linear correlation was observed for LS aBMD Z-score in individuals with fractures (Pearson’s correlation coefficient: − 0.8520, n = 375, nine studies; Fig. 3). After adjusting for sample size using the generalized linear mixed model (GLMM) approach, it was observed that for every 1-year increase in mean age, there was a 0.33 SDS decrease in LS aBMD Z-score for individuals with fractures (95% CI [− 0.38, − 0.28]; p < 0.001, Fig. 3).

Fig. 3.

LS aBMD Z-scores in individuals with and without fractures. aBMD areal bone mineral density, BMD bone mineral density, GLMM generalized linear mixed model, LS lumbar spine. Orange dots represent DMD participants with fractures and blue dots without fractures

Additionally, there was no significant difference in the mean age of individuals with and without fractures (Table 4). However, those individuals with fractures had lower mean aBMD Z-scores compared with individuals without fractures, with a significant difference in the LS aBMD Z-score (Δ = 0.51 SDS, p = 0.011) and total body less head (TBLH) BMD Z-score (Δ = 1.06 SDS, p < 0.001; Table 4) observed.

Table 4.

BMD Z-scores for DMD individuals with and without fractures at different anatomical sites

Characteristic	With fractures N = 198, 9 studies	Without fractures N = 190, 7 studies	Δ BMD Z-score	p value
Age, mean (SD)	10.49 (2.77)	10.84 (3.28)	-	0.579
LS aBMD Z-score, mean (SD)	− 2.18 (1.08)	− 1.67 (1.08)	0.51	0.011
TBLH BMD Z-score, mean (SD)	− 2.96 (1.37)	− 1.90 (1.37)	1.06	< 0.001

aBMD areal BMD, BMD bone mineral density, DMD Duchenne muscular dystrophy, LS lumbar spine, SD standard deviation, TBLH total body less head

Association between percentage of individuals with history of fractures with age, GC treatment duration, and LS aBMD Z-score

The relationship between the percentage of individuals with history of fractures, including long bone and VFs, and other disease variables was assessed (Table SI5). Age, GC treatment duration, and LS aBMD Z-score were all significantly associated with the percentage of individuals who have a history of fractures. For every 1-year increase in age and GC treatment duration, the percentage of individuals with fractures increased by 7.8% (95% CI [5.4, 10.2]; p < 0.001, n = 1291, 24 studies) and 8.3% (95% CI [− 0.1, 17.9]; p < 0.001, n = 799, 12 studies), respectively (Fig. 4a, b). Importantly, for every 1-SDS decrease in the LS aBMD Z-score, the percentage of individuals with fractures increased by 21% (95% CI [7.3, 34.8]; p < 0.01, n = 1066, 17 studies; Fig. 4c).

Fig. 4.

Association of different variables with the percentage of individuals with a history of fractures. a Percentage of individuals with history of fractures plotted against mean age of individuals with DMD. b The percentage of individuals with history of fractures plotted against mean GC treatment duration. c The percentage of individuals with a history of fractures plotted against mean LS aBMD Z-score. aBMD areal BMD, BMD bone mineral density, GC glucocorticoid, GLMM generalized linear mixed model, LS lumbar spine, pat patient, SDS standard deviation score

Detailed analysis of long bone and VFs in individuals with DMD (as reported in the records of this SLR) is provided in the supplementary information.

Discussion

This SLR and analyses were conducted to understand the natural history of bone fragility in individuals living with DMD, focusing on BMD, and the factors that affect its progression. Understanding BMD Z-score trajectories and fragility fractures among individuals with DMD is important to help monitor disease progression and implement potential preventive measures for individuals considered at risk or already suffering from the clinical manifestations of bone fragility.

The LS is by far the most common anatomical site to measure BMD in DMD, and LS aBMD has been the most frequently used method to report the Z-score. Nevertheless, BMAD and vBMD are being recognized as more precise methods for boys with DMD given the growth delay and osteotoxic effect induced by GCs, but the abundance of BMAD and vBMD data in the literature is still scarce (although steadily increasing), as shown in this SLR. However, given that the decline in BMD Z-scores is so prominent in DMD, the risk of fractures so high, and the fact that height Z-score decline, aBMD Z-score decline, and GC exposure are all co-linearly related, it is conceivable that aBMD bone size correction in DMD may be less important than in cohorts with less compelling relationships. On the other hand, greater rates of LS aBMD versus BMAD decline suggest part of the change may be due to the longitudinal growth impairment on GC. The need for appropriate aBMD size adjustments in DMD persists in order to ensure accurate descriptions of individual and aggregate patients’ BMD changes over time.

The analyses conducted found a strong association between LS aBMD Z-scores increasing age, and GC treatment duration in individuals with DMD. In this analysis, the reduction in LS aBMD Z-scores significantly accelerated at the age of 11 years, which is near to the mean age of LoA, a milestone that is also strongly influenced by age and motor function decline [64]. In addition, the rate of decline of LS aBMD Z-score in people with DMD who experienced fractures was faster than that observed in the analysis including all individuals (regardless of fracture status), suggesting that those individuals who have experienced fractures may be undergoing a different and more accelerated disease trajectory. It would be interesting to correlate this with motor function outcomes and determine whether those individuals experiencing a steeper decline in motor function have a history of early fractures. Unfortunately, North Star Ambulatory Assessment and/or timed function tests were not reported in most of the records included in this SLR in order to assess this correlation.

Furthermore, individuals older than 11 years of age, who are close to be or have started to be non-ambulatory, experienced a faster decline in LS aBMD Z-scores, more than tripling the rate of decline seen in younger individuals (5–10 years of age), who are mostly ambulatory. Crabtree et al. described this phenomenon, suggesting that the trabecular bone is highly affected after losing ambulation despite equivalent levels of GC exposure [65]. Lower LS aBMD Z-scores were observed in those individuals with versus without fractures. Harcke et al. showed that the occurrence of fractures in individuals with DMD was more than doubled when LS, LDF, or total body aBMD Z-scores were below − 2 but still above − 3, compared with aBMD Z-scores closer to the normal value (i.e., > − 2) [9]. There was a further increase in fractures when these aBMD Z-scores dropped below − 3 [9]. Additionally, a recent meta-analysis of randomized controlled trials assessing osteoporosis therapies in adults (males and females, in the general population) at increased risk of osteoporotic fractures by Black et al. suggested that treatment-related improvements in BMD were strongly associated with fracture reductions, irrespective of the mechanisms of action of the treatment being studied [66].

Moreover, in this SLR, similar declines in aBMD Z-scores were observed for measurements including TBLH, PF, and LDF, with LDF BMD Z-scores being more affected than any other anatomical location at any given age and showing the fastest rate of decline. This may partially explain why femur fractures in general, and distal femur fractures in particular, are among the most commonly observed long bone fractures in individuals with DMD [27]. Panicucci et al. reported that boys with DMD with fractures displayed a 1.5-fold higher decline in total body BMD Z-score/year [67]. Additionally, in the SPITFIRE trial (NCT03039686), the TBLH BMD Z-score showed a rate of decline after 1 year in line with the results obtained in this SLR (− 0.29 vs − 0.26 SDS/year, respectively), showing consistency and reliability of this analysis (Roche data on file).

Several other factors could influence BMD Z-scores, as identified in this SLR. GC exposure is known to have negative impacts on bone health and osteoporosis [4]. In this study, a strong association was observed between GC treatment duration and LS aBMD Z-score. However, we acknowledge that the cumulative dose of GCs may play a bigger role than the treatment duration, but unfortunately, this information was unavailable in most of the records included in this analysis. Moreover, our data suggest that prolonged exposure to GC treatment, despite its therapeutic benefits in managing DMD, correlates with lower LS aBMD Z-scores and a higher percentage of individuals with fractures, particularly VFs. This observation aligns with the fact that VFs are mainly driven by the use of high-dose GCs due to their osteotoxic effects on trabecular-rich bone.

Overall, preserving BMD in individuals with DMD is crucial in order to minimize LoA due to long bone fractures and other problems such as spine deformities due to VFs, which can significantly lower quality of life and create further complications. It is noteworthy that in DMD fractures can be observed even from birth and with BMD values “closer to normal”. One of the reasons that VFs can occur with apparently normal bone density is due to DXA’s inability to separate cortical and trabecular bone and consequently not detect the small changes in trabecular bone density which result in reduced bone strength. Thus, measuring vBMD by QCT could solve this issue by providing a true vBMD measuring only trabecular bone changes.

In light of these findings, we propose a trajectory roadmap of bone strength in individuals with DMD, illustrated in Fig. 5, which outlines critical milestones in DMD related to BMD. This road map highlights the early and aggressive nature of the osteoporosis in this context, underscoring the need for prevention strategies that are implemented early in life prior to first-ever fractures given the significant sequelae of fractures in this population (including permanent, premature LoA, and fat embolism syndrome).

Fig. 5.

Bone fragility trajectory in DMD. aBMD areal BMD, BMD bone mineral density, DMD Duchenne muscular dystrophy, GC glucocorticoid, LoA loss of ambulation, LS lumbar spine, VF vertebral fracture. Proposed timeline of key events in the bone fragility trajectory of individuals with DMD. ^aFractures can happen at any time in the course of the disease, but these are the mean ages reported in the literature. VFs can be observed within the first year of starting GCs

Limitations

A limitation of this SLR is that not all individuals included in the studies were treated with daily GCs, although studies of untreated individuals or individuals receiving other regimens (i.e., intermittent; explicitly mentioned in the reports) were excluded. Similarly, some individuals may have been treated with bisphosphonates and/or testosterone, but these details were not always reported, particularly for the older age groups included in the analysis. Nevertheless, we assumed bisphosphonate and testosterone naivete unless the manuscripts explicitly stated otherwise (underscoring the importance of reporting relevant details in a given population). There could also be other unreported confounding factors that could influence the results, such as pubertal stage, calcium and vitamin D supplementation, overall nutritional status, other concomitant medications, and physiotherapy regimens. In addition, none of the studies included herein studied the impact of vamorolone on bone strength.

There were only a limited number of studies reporting BMAD and vBMD, which could have influenced the significance of the results. The included older population (14–15 years old) in the SLR might not accurately reflect the late-ambulatory and non-ambulatory DMD populations due to a limited number of studies in this age range. It should also be noted that factors beyond BMD, such as small bone cross-sectional diameter (due to reductions in periosteal circumference), also contribute to bone fragility in DMD [68]. Given the paucity of data on this endpoint in clinical trials, we were not able to formally assess the relationship between long bone cross-sectional diameter and fracture risk, an opportunity for further investigation in future studies.

Given that not all the variables, such as age, BMD Z-scores, GC duration, and fracture data (among others) were reported across all records included, a multivariate analysis could not be performed with sufficient data to ensure reliable results.

Conclusions

To our knowledge, this is the first SLR investigating BMD Z-score trajectories in individuals with DMD receiving GCs. This study highlights the association between age and GC treatment duration, with a decline in LS aBMD Z-score evident in the first decade of life. These declines in BMD Z-score appear to play an important role in the increased risk of fractures in individuals with DMD, which is a high unmet medical need in this population.

A consistent decline in BMD accrual was observed in this analysis in individuals with DMD aged 5–15 years, which accelerated after 11 years of age as motor function declined on a pathway to LoA. Importantly, LDF BMD Z-score showed the lowest values overall at any given age and the highest rate of decline. GC treatment duration was the strongest factor affecting LS aBMD Z-score in individuals with DMD. The findings from this SLR emphasize the critical need for novel therapeutic strategies targeting bone fragility in DMD to improve bone health and minimize fracture burden. This is particularly true as efforts to effectively treat DMD by starting GC therapy at younger ages than the average reported in this trial are operationalized in contemporary times.

This comprehensive analysis of the natural history of BMD and the factors that influence it in DMD facilitates a better understanding of osteoporosis progression to support the improvement of bone strength among individuals with DMD.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This study was funded by F. Hoffmann-La Roche Ltd, Basel, Switzerland.

Data Availability

The data included in this article are available in the article and in the corresponding online supplementary material.

Declarations

Ethics approval

For this type of study, formal consent is not required.

Conflicts of interest

CDF and MG are employees of F. Hoffmann-La Roche Ltd with potential stock options. APM is an employee of Roche Products Ltd with potential stock options. YC is an employee of Genentech with potential stock options. CW has provided a consultancy role to F. Hoffmann-La Roche Ltd, PTC Therapeutics, and Pfizer. HJM has provided a consultancy role to F. Hoffmann-La Roche Ltd, Novartis, Pfizer, Regenxbio and Solid Biosciences. HJM’s institution has received research funding from F. Hoffmann-La Roche Ltd, Novartis, Biogen, Dyne, PepGen, Sarepta, Italfarmaco, and Reveragen. EM has received fees from AveXis, Biogen, and F. Hoffmann-La Roche Ltd. NC has nothing to disclose. LMW has participated in clinical trials with ReveraGen Biopharma, Catabasis, and Edgewise Therapeutics and has provided a consultancy role to Santhera, Catalyst, F. Hoffmann-La Roche Ltd, and PTC Therapeutics, with funds to Dr Ward’s institution.