Essentially, Poehlman forged data for nearly a decade in longitudinal studies of metabolic changes in women going through menopause. In general, it seems as though he made his "data" support the hypotheses most scientists already believed were correct, and thus gained prominence and funding. Of course, as an NIH researcher pointed out, forging data was hardly necessary for him to get grants:
Had Poehlman done the hard work in the menopause study, had he followed through on the unexpected data on lipids, he might have eventually achieved the same status. That is the tragic twist. "Much is revealed when studies don't go the way you expect them to," says Sally Jean Rockey of the N.I.H. "Because Poehlman was working in critical areas for lots of people, what he learned would have been important either way."That's one of the key points that I wish my students would understand. Almost everyone who walks into my lab at the beginning of the semester believes that if the data they collect don't support their initial hypotheses, then the data must be wrong1. Instead, that's when things get interesting.
1 I have a strong suspicion that chemistry instructors are largely responsible for this belief, though I have no evidence for that.