Do patients with high historic peak panel-reactive antibodies (PRA) remain high risk if their PRA levels fall before transplantation? We examined retrospectively 406 first and repeat kidney recipients with a peak PRA of >50%, who were transplanted from our center between January 1990 and December 2001. Univariate analysis by log-rank test was performed for variables that affect graft survival. The factors tested included current PRA, peak PRA, difference between peak and current PRA (DeltaPRA), HLA mismatch, gender, age, transplant number, and donor source. Receiver operator characteristic curves (ROC) were generated to obtain the best cutpoints for current PRA and DeltaPRA. Current PRA (P < .0001), peak PRA (P = .0004), and DeltaPRA (P = .0015) were significant predictors by univariate analysis. However, in a multivariate model, peak PRA was not significant. Current PRA (P < .0001) was significantly associated with graft survival, while DeltaPRA showed a strong trend to significance (P = .05). Current PRA of <26% and DeltaPRA of >37% were the best cutpoints for separating good and poor outcomes. This study shows that current PRA and DeltaPRA impact on graft survival in highly sensitized (>50%) patients. Sensitized patients with peak PRA >50% who subsequently have a drop in PRA to <26% are at lower risk of graft loss than those with a persistently high PRA. A fall in peak PRA of >37% at the time of transplant appears to be of benefit only in those patients who achieve a current PRA of <26%.