Merck knew about Viox' problems from the research they suppressed.
The data matters mainly to retrospectively determine how bad the situation was and looks like it was really bad.
What's interesting here though is that this is not a small signal, precisely because it shows up against the background without any correlation work at all. FTA:
> "We find the largest rise in American mortality rates [in the last 15 years] occurred in 1999, the year Vioxx was introduced, while the largest drop occurred in 2004, the year it was withdrawn," says Unz.
(Source article for that graph: http://www.cdc.gov/nchs/data/databriefs/db88.htm -- would that they distributed the actual datasets behind the graphs!)
EDIT: URL removed. The link I included appear to be written by an idiot. Sorry.
> A CDC review of mortality patterns in 1993 states: “the decline in life expectancy likely reflects increases in death rates for chronic diseases during the two influenza outbreaks in 1993.” [Morbidity Mortality Weekly 45 (08), 161-64, March 1, 1996] Some increases in chronic disease (diabetes, heart disease, COPD) were “the result of the two influenza epidemics of 1993,” said the report. [Monthly Vital Statistics Report, Volume 44, No. 7(S), page 9, Feb. 29, 1996]
Apologies -- I'd misread the OP to say that it had been the largest jump in 15 years -- meaning the 15 years prior to 1999.
http://www.cdc.gov/nchs/data/databriefs/db88_fig1.png
If you know where to look, there is indeed a small spike from 2000-2005, corresponding with the availability of Vioxx. The rise in 2000 is barely noticeable, but the drop in 2005 is fairly striking. This makes sense: the drug would have been taken up gradually by the market, but its cessation was virtually overnight.
The reason this would be difficult to spot, however, is because death rates did not really increase. Rather, they stayed more or less constant, when they might otherwise have gradually decreased. In hindsight, this looks plausible -- but at the time, saying that death rates would be decreasing were it not for Vioxx would have been an extremely difficult counterfactual argument to make.