Biased, maybe. But the sample sizes are pretty big, whether "bigness" refers either to absolute numbers or to being big enough such that whatever they're estimating will come pretty close to the true value of that variable for similar populations. I mean, the smallest sample is 70k plus.

No, but as always they trot out the relative numbers.

    > "25% greater risk of death, compared with H2 blockers"

Whether that means anything to me depends on what my odds of dying are on H2 blockers. If it's 1 in 3 then I care. If it's 1 in 3,000,000 then I don't.