It is well known that the disease course in Duchenne muscular dystrophy (DMD) patients is heterogeneous, varying from patient to patient. Such heterogeneity is also seen in the golden retriever muscular dystrophy (GRMD) dogs. In this context, identifying predictive markers of the clinical evolution would allow a better management of the model during therapeutic trials. This study aimed to identify biomarkers, able to predict the evolution of disease in the GRMD dogs towards one of the two clearly distinct clinical forms: the severe form (SF, loss of ambulation before the age of 6 months), and the moderate form (MF, no loss of ambulation). The proportion of circulating T-lymphocytes expressing high membrane levels of CD49d (the alpha-chain of the integrin VLA-4, previously defined by our group as a progression biomarker of human DMD), was assessed in seventeen 2 month-old GRMD dogs (5 SF, 12 MF). Clinical and functional tests (motor score (10 SF, 14 MF), and gait analysis using accelerometry at 2 months (9 SF, 10 MF), force measurement at 4 months (5 SF, 11 MF)) were also assessed to predict the two clinical forms of disease progression. The proportion of circulating CD4 + VLA4hi T-lymphocytes was significantly increased (p = 0.002) in SF dogs. Additionally, two month-old SF dogs were not able to maintain as frequent locomotor cycles as MF dogs (p = 0.007). A stride frequency lower than 2.3 s?1 was shown to be able to predict SF, with a Sp of 90% and a Se of 78%. This particular contractile behaviour probably reflects a more impaired ionic homeostasis in SF dogs. Accordingly, 100 ms post-tetanic relaxation level lower than 43.2%, was able to predict SF with a Sp of 91% and a Se of 100%. These predictive markers offer satisfying levels of specificity and sensitivity, which could be increased by combining the different markers, to reliably predict SF or MF. This combined strategy may be useful for better evaluating therapeutic trials.