In this study, we evaluated the effect of imprecision in patient repositioning encountered in real life on multiple sclerosis (MS) lesion volumes measured from MRIs. We also evaluated two putative methods for reducing the variability in these lesion volume measurements: first, a reduction of slice thickness (from the conventional 5 mm to 3 mm) and second, the application of a new repositioning technique based on the use of head immobilization shells. We evaluated the errors in lesion volume by scanning 10 patients a total of four times using the two slice thicknesses and two repositioning methods (conventional and using a head immobilization shell). The mean absolute percentage difference between two corresponding scans was 6.8% (range, 1.24 to 11%) using conventional slice thickness and repositioning, 4.1% (range, 0.7 to 5.56%) using conventional slice thickness and head immobilization shells, 2.6% (range, 0.8 to 6.66%) using the conventional repositioning technique and 3-mm slice thickness, and 1.4% (range, 0.2 to 6.14%) using slice thickness of 3 mm and head immobilization shells. These mean absolute differences were significantly different (p = 0.0008). Our results indicate that the effect of repositioning errors of the order of those that can be encountered in the daily life situation of clinical trials affects significantly lesion load measurements in MS and that the combined use of thinner slices and more accurate repositioning techniques can markedly improve the reproducibility of such measurements.