A reanalysis of a longitudinal scleroderma clinical trial using non-ignorable missingness models