A total of 57 PFS events were
required. All analyses were carried out on an intent-to-treat basis. PFS curves
were derived using the Kaplan-Meier method. PFS was analyzed using a Cox
proportional hazards model with factors for treatment (olaparib v PLD),
BRCA1 or BRCA2 status, and platinum sensitivity (sensitive or resistant/
refractory). An HR of 0.56 or better would be required to be observed with 57
events to achieve a P.02 (one-sided). In the event of nonsignificance in the
combined comparison, each olaparib dose was to be compared separately with
PLDusing a closed testing procedure and type I error rate of 0.005 (one-sided).
Theoverall type I error for declaring a statistically significant difference in favor
of olaparib was to be no more than 2.5% (one-sided). ORR was compared
between olaparib (combined and separately) and PLD using logistic regression,
adjusting for the same set of factors as for PFS. Effect of treatment was
estimated using the adjusted odds ratio. Changes from baseline in target tumor
size were assessed using analysis of covariance. OS was assessed using a Cox