We studied the effects of LTBP4 and SPP1 polymorphisms on age at loss of ambulation (LoA) in a multiethnic Duchenne muscular dystrophy cohort (DMD).We genotyped SPP1 rs28357094 and LTBP4 haplotype in 283/340 participants in the Cooperative International Neuromuscular Research Group Duchenne Natural History Study (CINRG-DNHS). Median ages at LoA were compared by Kaplan-Meier analysis and log-rank test. We controlled polymorphism analyses for concurrent effects of glucocorticoid corticosteroid (GC) treatment (time-varying Cox regression), and for population stratification (multidimensional scaling of genome-wide markers).Hispanic and South Asian participants (n = 18, 41) lost ambulation 2.7 and 2 years earlier than Caucasian (p = 0.003, <0.001). The TG/GG genotype at SPP1 rs28357094 was associated to 1.2-year earlier median LoA (p = 0.048). This difference was greater (1.9 years, p = 0.038) in GC-treated participants, whereas no difference was observed in untreated. Cox regression confirmed a significant effect of SPP1 genotype in GC-treated participants (HR 1.61, p = 0.016). LTBP4 genotype showed a direction of association with age at LoA as previously reported, but not statistically significant. After controlling for population stratification, we confirmed a strong effect of LTBP4 genotype in Caucasians (2.4 years, p = 0.024). Median age at LoA with the protective LTBP4 genotype in this cohort was 15.0 years, 16.0 if GC-treated.In conclusion, SPP1 rs28357094 acts as a pharmacodynamic biomarker of GC response, and LTBP4 haplotype modifies age at LoA in the CINRG-DNHS cohort. Adjustment for GC treatment and population stratification appears crucial in assessing genetic modifiers in DMD. This article is protected by copyright. All rights reserved.