Additional detail added regarding the exclusion of previously irradiated lesions from the analysis, as the effects of radiotherapy on tumours may manifest more than 6 weeks after completion of radiotherapy. Therefore inclusion of previously irradiated lesions in the analysis may have impacted the primary outcome variable of CBR. Also added further detail around concomitant use of vaginal creams and vaginal rings during the study. New evidence had come to light regarding the significance of the systemic absorption of oestrogens from topical vaginal preparations used in the treatment of menopausal symptoms (atrophic vaginitis) in patients taking AIs (Kendall et al 2006). The authors were of the opinion that the use of topical vaginal oestrogen-containing creams is contraindicated in patients taking AIs. On this basis, and also because the use of these creams had the potential to bias the study against the AI arm, the study restrictions were clarified to specifically preclude the use of oestrogen-containing vaginal creams or other oestrogen containing topical preparations. However given the distressing nature of atrophic vaginitis, it was felt necessary from an ethical point of view to allow the use of controlled-release oestrogen containing vaginal rings (e.g., Estring®), as a last resort only, pending any updated advice, and only at the discretion of the investigator.

At the DCO for the primary analysis, patients on fulvestrant 500 mg had a 60% longer TTP compared to patients on anastrozole 1 mg. The data maturity in terms of the proportion of patients with progression events was approximately 30%. A more mature analysis, planned for when approximately 75% of patients had discontinued study treatment, allowed a more meaningful interpretation of this secondary endpoint. Since patients were not followed within the study according to formal scheduled RECIST assessments from the time of primary DCO onwards, a “time to treatment failure” (TTF) analysis was planned where treatment failure was determined according to the investigator’s opinion. This amendment also clarified the timing for the final analysis, and also documented that, for subjects still on study treatment, only serious adverse event data would be collected following the final analysis.

Following a statistically significant benefit (time to progression hazard ratio = 0.66, 95% confidence interval (0.47, 0.92), p=0.01) in favour of fulvestrant, seen in the follow-up analysis of time to progression (TTP), it was decided to amend the protocol to investigate if the benefit observed for TTP translates into an overall survival (OS) benefit. The amendment covered the collection of OS data and detailed the performance of an OS analysis. It clarified that the “time to treatment failure analysis” (TTF) was not the final analysis for this study.

The amendment covered the collection of overall survival (OS) data and described the change in timing for the final OS analysis. Significant decline in rate of death had been observed during the later stages of overall survival follow-up. This rate had become very unpredictable as the CSP amendment 3 was looking for events in a small group of patients. It was possible that 75% deaths would not be reached for several years, which was not considered a reasonable timeframe for obtaining information on OS. Performing the analysis at approximately 65% rather than 75% maturity did not make a large difference to the outcome, in terms of the largest observed HR which would achieve statistical significance (p<0.05); 0.71 and 0.73 for 65% and 75% respectively.