Ah, that one by that American company using antibodies. Tough things, antibodies. They're very specific.
Animal tests aren't good on antibodies, because these antibody proteins are specific. They're very specific. You change the shape of their targets just slightly and you can alter their affect and efficiency.
The drug was TGN 1412 and was more of an antibody. It worked by enhancing an action by binding to a T-cell's CD28 receptor (It's been four years since I last did immunology so I can't remember what that one does). That's a dangerous thing to do, you know, enhancing bodily functions. You could overstimulate something.
These men were healthy, after all. They had perfectly fine immune systems. You boost a good immune system and you get an overactive one, which would explain why they reacted so badly.
http://news.bbc.co.uk/1/hi/health/4817178.stm
It has even been suggested in a New Scientist issue that they were given a potentially nasty underdose. After all, they were given 0.2% of the dose considered safe in animal experiments. If TGN1412 has a stronger affinity for "conventional" T-cells than for regulatory T cells, a low dose may have been enough to stimulate the former but not to stimulate the protective regulatory T cells.
Inevitably, we don't know what went wrong. After this was a new "drug" targeting a never before targeted receptor.