Political numbers whiz Nate Silver gives a long and poll-technical presentation about why it is perfectly likely, given some current polling, that they could lose up to 51 seats.
The heart of his analysis:
the House popular vote (a tabulation of the actual votes all around the country) and the generic ballot (an abstraction in the form of a poll) are not the same thing—and the difference usually tends to work to Democrats' detriment. Although analysts debate the precise magnitude of the difference, on average the generic ballot has overestimated the Democrats' performance in the popular vote by 3.4 points since 1992. If the pattern holds, that means that a 2.3-point deficit in generic ballot polls [which we see now in Real Clear Politics polling] would translate to a 5.7 point deficit in the popular vote—which works out to a loss of 51 seats, according to our regression model.
On Nate Silver's history of political prognostication success.