7. Statistical power
7.1 Plausible reductions in breast cancer mortality
In 2012, the UK panel on breast cancer screening estimated an absolute reduction in breast cancer mortality of 0.6 per 1000 women per routine 3-yearly invitation in the age range 50-70 (total 4.3 per 1000 for 7 such invitations).1 The probability that mammography will detect a cancer is, however, only about one-third as great at ages 47-49 as at ages 71-73. This, together with the panel’s estimate, would suggest absolute reductions of about 0.3 per 1000 younger invitees and 0.9 per 1000 older invitees.
UK mortality from breast cancer in middle age may well, however, be about one-third lower in the 2020s than it was in the 2000s. So, reductions in mortality of only about 0.2 per 1000 younger invitees and 0.6 per 1000 older invitees can plausibly be expected during follow-up in AgeX.
The plausible reductions outlined above provide the basis for the AgeX power calculations, and adjustments for adherence (Section 6.3.1) have little or no effect on this power (as they do not, in expectation, alter the absolute difference between the two treatment groups in numbers of relevant breast cancer deaths, or in the variance of this difference).
However, as entry into AgeX has already ended and the main analyses population has already been defined (Table 1), the statistical power depends entirely on how big the effect on breast cancer mortality of just one additional screening visit will actually be, which will not be known until after the final results emerge.
7.2 Power calculations for the final primary analyses of breast cancer mortality
The statistical power of the final primary analysis, with follow-up of mortality to 12.2031, will be greater than the statistical power of the interim primary analysis with follow-up only to 12.2026, unless there is little additional benefit after 2026. The expectation of an absolute reduction of 0.2 per 1000 younger invitees is consistent with the hypothetical numbers in Table 2, which underlie this SAP.
In Table 2, the absolute reduction of 0.2 breast cancer deaths per 1000 invitees is driven by the 20% reduction after age 55 from a cancer diagnosed < 4 years after randomisation, involving 400 vs 500 deaths expected before age 60 and 400 vs 500 deaths after age 60. If a total of 800 vs 1000 deaths is expected, this gives a 98% chance of achieving P < 0.01 and a 92% chance of achieving P <0.001.
The power for the final analysis, with follow-up to 12.2031, to detect an absolute reduction of 0.6 breast cancer deaths per 1000 older invitees would be at least as great, for although there are only half as many older as younger women in AgeX the older women who will be included in the main analyses population have thus far (in analyses blind to the treatment allocation) had about as many breast cancer deaths in total as the younger women.