Annex 1: Changes to the previously described study design and follow-up

Partly to increase statistical power, randomisation of older women between more than one additional screening invitations and no additional invitations was originally envisaged, but the additional work required was not practicable for the screening program, so the AgeX trial involves only one additional invitation.

The routine assignment of underlying causes of death to breast cancer (ICD-10 C50 or ICD-11 2C6) or to other 3-character ICD categories by the Office of National Statistics will be accepted without further investigation. It was originally envisaged that all possible deaths from breast cancer would be reviewed, but this has not been attempted for any deaths so far, and will not be. For, attempts by the investigators to refine ONS cause of death assignments would have been expensive, time consuming, and incomplete, would have had little effect on statistical power, and could have engendered unjustified concerns about possible effects of investigator biases.

To increase sensitivity of the primary analyses and make the subsidiary analyses relevant to them, these analyses were all to be limited to the women of appropriate age who could be linked to national statistics, had no prior cancer or breast disease, and were likely to accept an invitation if randomly allocated to receive one. The details of these analyses (which are given in the statistical analysis plan) could not, however, be finalised until after recruitment had ended and linkage to all relevant national databases had been attempted. It was done blind to any mortality analyses.

The protocol for the randomisation phase of the trial envisaged that, when sufficient follow-up was available, subdivision of the findings for breast cancer mortality by 5-year groups of time since randomisation would supersede the previously planned subdivision by whether the deaths were before or after age 60 for younger women (or 80 for older women). Because the age range of randomisation is extremely narrow, the present subdivision of mortality by 5-year age groups achieves both purposes, and has been adopted blind to any mortality analyses.

Interim and final analyses will report mortality to 12.2026 (with the first mortality report due in 2027-28) and then to the 12.2031 (with the final mortality report due in 2032-33). The final mortality report, will be accompanied by health economic analyses.