Predictors for relapse in a double-blind randomized placebo-controlled discontinuation study of remitted first-episode psychosis patients