3. Study methods
3.1 Trial design
AgeX comprises two large cluster-randomised trials (one in older and one in younger women), each with about 20,000 small clusters of women. Within each trial, each cluster was randomly allocated (as if on the toss of a coin) to receive, or not, one additional invitation to mammographic screening.
3.2 Randomisation
The breast screening programme (BSP) in England is delivered by around 80 breast screening units (BSUs); the exact number fluctuates. They could not join AgeX until they were using digital mammography. AgeX began in 2009 with a pilot study of the acceptability of cluster randomisation in 5 BSUs. By 2013 it had been extended to around 60 BSUs (after which the number of participating BSUs remained approximately stable; all were in England), and it randomised about 0.5 million women annually from then until March 2020.
3.2.1 Definition of age as a whole number
Within the BSP, a national database is used to create screening invitation batches every few weeks for each BSU. For each of the BSUs recruiting into AgeX, typical invitation batches included a few hundred names spanning ages 47-73 (which the BSP defined in whole years by year of birth until mid-study, and by exact date of birth thereafter: Annex 3 to study protocol). Some batches also included small numbers of women aged 45-46, classified in this SAP as 47, or aged 74-75, classified in this SAP as 73.
3.2.2 Definition of clusters for randomisation
The intent was that those aged 50-70 would not be newly entering AgeX and would instead all be sent routine screening invitations (regardless of whether they were already in AgeX), but that the two “clusters” of younger or older women would both enter the study, with either the younger cluster or the older cluster invited for screening. The program that generated the batch defined which women were newly entering AgeX, and decided randomly, in a 50:50 ratio, which of the two clusters would be invitees and which would be controls. (The older and the younger cluster do not get compared.)
3.3 Analysis as two separate trials
The findings from AgeX will be monitored, analysed and reported as two entirely separate trials. One is a trial among younger women (randomised at age 47-49 to be invitees or controls) of the effect of one additional screening about 3 years before routine screening would normally have begun. The other is a trial among older women (randomised at age 71-73 to be invitees or controls) of the effects of one additional screening about 3 years after routine screening would normally have ended. Thus, the younger clusters in AgeX get compared only with each other, and never with any of the older clusters; likewise, the older clusters get compared only with each other.
3.4 Sample size
AgeX did not have a pre-determined sample size. Instead, by design, randomisation could have continued for as long as substantial uncertainty about the results persisted, unless the data monitoring and ethics committee advised otherwise.
From 2009-20, around 20,000 clusters containing 2.8 million younger women were randomly allocated to be invited or not for one additional screening. Likewise, around 20,000 clusters containing 1.7 million older women were randomly allocated to be invited or not for one additional screening.
The main analyses population will include only the 2 million younger women and 1 million older women who had been eligible and linkable to NHS England datasets, with no record there of prior cancer or breast disease, and who were considered likely to attend screening if invited (based on uptake of their previous breast or cervical screening invitation). These criteria depended entirely on information that could not have been affected by the random allocation. Details of those included and excluded are in Section 5.2. Those excluded will be analysed in appropriate ways (Section 6.8), but not as part of the main analyses.
3.5 Hypothesis framework
For each of the main analyses, the null hypothesis will be no effect of one additional screening.
3.6 Need for long-term follow-up of mortality
Long-term follow-up is necessary to assess the eventual effects of one additional screening on breast cancer incidence, breast cancer treatment, and, particularly, breast cancer mortality. As little effect on breast cancer mortality should be expected until years 5-9 and 10-14 after randomisation, prolonged follow-up of mortality will be required, first to 12.2026 and then to 12.2031.
3.7 Interim analyses and “alpha-spending”
Analyses of breast cancer incidence and treatment will be reported in 2025-26, before any unblinded analyses of mortality are reported, but electronic follow-up of breast cancer incidence and treatment will continue after that date.
Unless the DMEC recommends release of earlier mortality results, the only interim report on mortality will be that scheduled to be released in 2027-28, based on analyses of mortality to 12.2026, and the second and final report on mortality will be that scheduled to be released in 2032-33, based on analyses of mortality to 12.2031. The final report on mortality will be accompanied by updated analyses of breast cancer incidence and treatment, and by health economic analyses.
P-value calculations and confidence intervals in the final analyses will make no formal “alpha-spending” statistical allowance for the existence of previous analyses, as electronic follow-up will continue regardless of any interim findings.