In Pfizer's failed trial, there were about 130 deaths out of a high-risk patient population of 15,000 when the trial was stopped after one year. So that's under 1% annual incidence, and an even lower incidence in the placebo group. So it would be very hard to get any decent mortality statistics on a trial with less than 10k patient-years unless the effect was really dramatic.
In terms of your suggested surrogates, the torcetrapib trial showed slightly reduced atheroma (non-significant on primary endpoint, significant on one secondary endpoint) via IVUS on 1,200 patients after 2 years. So that wasn't a cheap magic bullet either. Increased fat in stools isn't a validated endpoint at all.
(I personally believe Torcetrapib failed because it induced slight hyperaldosteronism and also increased cortisol levels - those are not class effects).
The recent Niaspan trial that was stopped prematurely did show IMT effects via IVUS in a very much smaller population, so I guess if the drug really works you can see some effects in a smaller trial.
Peter