Table 3.

Conditional probabilities and predictive value of early treatment response at Week 1, Week 2, Week 3, Week 4 and Week 8 predicting ultimate treatment response at Week 12

Predictor Variable a	Early Responder, n (%)	Odds Ratio	P	Sensitivity (%)b	Specificity (%)c	PPV(%)d	NPV(%)e	Accuracy
Week 1	5/100 (5.0%)	1.21	0.83	9.1%	69.7%	9.1%	69.7%	54.5%
Week 2	18/100 (18.0%)	5.18	0.03	66.7%	72.1%	26.1%	93.6%	71.4%
Week 3	21/94 (22.3%)	5.73	0.01	66.7%	74.1%	34.8%	91.5%	71.4%
Week 4	28/89 (31.5%)	18.11	0.02	76.5%	88.0%	65.0%	90.7%	83.9%
Week 8	24/81 (29.6%)	19.36	0.008	73.9%	87.2%	73.9%	87.2%	82.9%

PPV Positive predictive value, NPV Negative predictive value

^*P-value< 0.05 statistically significant

Predictor Variable: ≧25% reduction in PANSS total scores

a Sensitivity x: correct identification of responders, e.g. if 100 patients responded

at endpoint, x (of these 100) patients were correctly identified earlier

b Specificity x: correct identification of non-responders, e.g. if 100 patients failed to respond at endpoint, x (of these 100) patients were correctly identified earlier

c PPV x: proportion of early responders who were responders at endpoint, e.g. if 100 patients were early responders, x (of 100) responded at endpoint

d NPV x: proportion of early non-responders who were non-responders at endpoint, e.g. if 100 patients were early non-responders, x (of 100) failed to respond at endpoint