It has nothing to do with cancer development, everything to do with finding the association between a risk factor and cancer development. Think of a biased coi

It has nothing to do with cancer development, everything to do with finding the association between a risk factor and cancer development. Think of a biased coin that tends to come up heads too often. If it does it every time, you don't need to flip it very many times to figure out beyond a reasonable doubt that it's biased. But if it comes up heads 6 times out of 10, you will have to flip it a lot to be sure that it's bias and not just chance. In medicine most associations are more like the latter case than the former. I'm speculating that the juice sample is like a too-short run of flips on a slightly/moderately biased coin that's and therefore can't show an association, and the soda sample is like an adequately long run of flips on a similar coin. The sugar in juice is about the same as the sugar in soda, so this seems like the most plausible explanation.