Male participant needed to make a minimum of 4 attempts at SI and Baseline IELTs were assessed by stop watch measurements for each SI attempt. IELT was the elapsed time from vaginal penetration until ejaculation. On-treatment IELT for each participant was calculated by taking the median IELT from all valid on-treatment IELT attempts. Geometric mean IELT for each treatment was compared using analysis of covariance (ANCOVA). LS mean values and two-sided p-values are from the general linear model ln(median IELT)= ln(Baseline IELT) + treatment + cluster. A step-down procedure was used to determine if the efficacy of on-demand GSK557296 was superior to placebo. First, the average of the geometric mean IELTs of the pooled 150mg and 50mg doses of GSK557296 was tested against placebo, and if significance was achieved with this global plateau trend test (p \< 0.05), then the simultaneous pair wise comparisons of the 150mg and 50mg doses to placebo would occur (each at an alpha level of 0.05).