As a first step, an initial latent growth model (LGM) was fitted in which the slopes and intercepts of the follow-up PDA values (months 3 to 27) were modeled, with baseline PDA was used as an exogenous predictor. Several alternative forms were considered in modeling the pattern of PDA over time. Among these were models containing both linear and quadratic components (χ2=106.95; df = 42; p < .01), a piecewise model in which slopes for PDA were analyzed separately for months 3 to 15 and months 18 to 27 (χ2=99.25; df = 42; p < .01), and a piecewise model in which slopes for PDA were analyzed for months 3 to 18 and 21 to 27 (χ2=115.49; df = 42; p < .01). A linear model yielded a fit that was comparable to the other models (χ2=121.18; df = 47; p < .01).
