To determine whether early manic symptom improvement predicts outcome in bipolar I disorder patients experiencing manic or mixed episodes.

Pooled data from two 3-week trials; asenapine (5 or 10 mg BID; n = 372), olanzapine (5-20 mg QD; n = 391), or placebo (n = 197). Early improvement (YMRS total score changes from baseline ≥15%, ≥20%, and ≥25%) was assessed at days 2, 4, and 7. Associations between early improvement and week 3 outcomes (YMRS response [≥50% total score reduction] and remission [total score ≤12]) were calculated using Fisher's exact tests; odds ratios classified their relative strength. Sensitivity (SN), specificity (SP), and positive (PPV) and negative (NPV) predictive values were calculated as previously described (J_Clin_Psychiatry_2009;70:344–353).

Early improvement was strongly associated with positive outcomes. the earliest positive associations across all cutoffs occurred with asenapine at day 2 (response, all P < 0.04; remission, all P < 0.007), olanzapine at day 4 for response (all P < 0.02) and day 2 for remission (all P < 0.002), and placebo on day 7 (response, all P < 0.003; remission, all P ≤ 0.0005). Odds ratios were higher for asenapine (1.8–9.1) than olanzapine (1.4–3.5) and placebo (1.3–8.0). Respective day 4 remission values for SN, SP, PPV, and NPV at the ≥ 15% cut-off were 80%, 58%, 48%, and 85% for asenapine; 76%, 43%, 49%, and 71% for olanzapine; and 50%, 67%, 31%, and 82% for placebo.

Early improvement was strongly associated with week 3 response and remission; high NPVs indicated little chance of stable remission in the absence of early improvement.