Prediction models for natural history trajectories can provide important benchmarks for the outcomes of patients receiving novel therapies, especially over multi-year periods for which placebo controls are not feasible. To this end, we developed and validated a prediction model for changes in the North Star Ambulatory Assessment (NSAA) total score using longitudinal natural history data from 24 ambulatory male Becker muscular dystrophy (BMD) patients at Leiden University Medical Center (LUMC). Median follow-up was 37.0 months (range 11.7–60.7). Mean baseline age was 39.4 (standard deviation 12.6) years and mean baseline NSAA score was 23.9 (10.6) points. Models were fit using multivariable longitudinal regression, considering different sets of predictors, with the final model selected based on the corrected Akaike information criterion. The best-performing prediction model included linear trajectories with slopes modified by baseline age, NSAA total score, 10-meter-walk/run velocity, four-stair-climb velocity and rise-from-supine velocity. This model explained 53% of variation in NSAA total score changes from baseline, with an average prediction error (root-mean squared error) of 2.05 NSAA units. For external validation, the prediction model was applied to published average baseline profiles and NSAA changes from baseline. Predicted mean NSAA changes were similar (-0.9 [predicted] vs -0.9 [observed]) as observed at 12 months in Bello et al. (2016) and indicated slightly less decline (-0.9 vs -1.3 at 9 months, and -1.9 vs -2.5 at 18 months) than observed in De Wel et al. (2024). While this model was trained on a relatively small number of subjects, and updates with additional training data are warranted, its performance on external data sources shows adequate predictive performance to benchmark and contextualize NSAA treatment outcomes over 2+ years in ambulatory BMD.