traj

Data are from a schizophrenia clinical trial comparing placebo, standard treatment (haloperidol), and experimental treatment (risperidone). There are 87 subjects assigned to each of the placebo and haloperidol arms and 168 assigned to the risperidone arm. The response is the Positive and Negative Syndrome Scale (PANSS), assessing psychotic symptoms. Higher scores indicate greater severity of psychotic symptoms. Measurements were taken at baseline, 1, 2, 4, 6, and 8 weeks and there is considerable attrition among subjects. The example illustrates how to fit a three-trajectory model to the PANSS scores with risperidone treatment as a risk factor. Dropout at each wave is modeled to depend on level and trend in PANSS scores as well as having received risperidone treatment. Output and graphs follow.

Model Estimates and Graphs

There is a group of subjects with high PANSS scores and no improvement (group 3, 14.1%); a group with more moderate severity with a 9 point decrease in PANSS at week 8 (group 2, 45.7%); and a group with the lowest severity and the most improvement, a 24 point decrease at week 8 (group 1, 40.2%). Risperidone treatment reduces the likelihood of being in the poorer treatment response groups (2.6 times less likely to be in group 2 vs. 1, 3.0 times less likely to be in group 3 vs, 1).