The last observation carried forward (LOCF) analysis overlooks the meaningfulness of dropout in clinical trials. For antipsychotic medication dropout is an important outcome since long-term treatment is often required and dropout may relate to lack of drug tolerability. OBJECTIVE: The current analysis applies the “pattern mixture” approach (Shih & Quan, 1997) in which a composite hypothesis is tested that consists of the probability that there is a difference in completion rates (d) between two drugs and the probability that there is a difference in efficacy of complete cases (e) [ p = p(d) x p(e) x (1- ln( p(d) x p(e) ]. METHODS: The pattern-mixture approach was applied to data from a 53-week randomized, open-label non-inferiority efficacy trial of risperidone long-acting injectable (RLAI) vs. olanzapine tablets (OLA) in treating schizophrenia (n=618) (data on file JNJ). RESULTS: LOCF had found a significant difference (p=.04) on percent of patients in each group who attained clinical improvement (20% improvement on PANSS total) favoring RLAI and no significance on difference in the continuous measure of change in PANSS total (p=0.83). Among completers there was a greater decline on change in PANSS total favoring RLAI (Ris –23.6(±14.4); Ola –21.9 (±18.0); 1-tailed p=0.105). 76% of the RLAI treated patients completed the trial as compared to 70% of the OLA treated patients (one-tailed p=0.087). Using the pattern mixture approach the probability for the combined hypothesis of a difference in efficacy in complete cases, and trial completion, was significant (p=0.05). On clinical improvement ,66.1% of RLAI group both completed the trial and improved as compared to 53.7% of the olanzapine group (Odds Ratio [95% Confidence interval]: 1.84 [1.20:2.82]). CONCLUSIONS: LOCF may not capture real-life, clinically important differences which can be captured by other approaches.