A recent poll shows Ben Carson ahead of Donald Trump in Iowa, which may be a temporary fluke or may signal that Carson will win the state's primary with Trump in a close second.
Even in the worst-case scenario where Carson wins Iowa, would that spell doom for Trump's shot at the nomination? No: the results of the Iowa primary are a poor predictor for what happens nationally. The primaries in New Hampshire and South Carolina are far more accurate predictors, and there Trump enjoys a comfortable lead.
From Wikipedia, here are the past results of the Iowa Republican primaries where an incumbent was not running. Winners in Iowa are listed first, winners of the nomination in bold.
2012 - Rick Santorum (25%), Mitt Romney (25%), Ron Paul (21%), Newt Gingrich (13%), Rick Perry (10%), Michele Bachmann (5%), and Jon Huntsman (0.6%)
2008 – Mike Huckabee (34%), Mitt Romney (25%), Fred Thompson (13%), John McCain (13%), Ron Paul (10%), Rudy Giuliani (4%), and Duncan Hunter (1%)
2000 – George W. Bush (41%), Steve Forbes (31%), Alan Keyes (14%), Gary Bauer (9%), John McCain (5%), and Orrin Hatch (1%)
1996 – Bob Dole (26%), Pat Buchanan (23%), Lamar Alexander (18%), Steve Forbes (10%), Phil Gramm (9%), Alan Keyes (7%), Richard Lugar (4%), and Morry Taylor (1%)
1988 – Bob Dole (37%), Pat Robertson (25%), George H. W. Bush (19%), Jack Kemp (11%), and Pete DuPont (7%)
1980 – George H. W. Bush (32%), Ronald Reagan (30%), Howard Baker (15%), John Connally (9%), Phil Crane (7%), John B. Anderson (4%), and Bob Dole (2%)
1976 – Gerald Ford (45%) and Ronald Reagan (43%)
So, Iowa has only predicted the nominee in 3 out of 7 cases. Not exactly a bellwether. They routinely pick soft-spoken candidates who are nevertheless destined not to secure the party's nomination. In recent primaries, they have favored candidates with strong evangelical appeal, who do not go over well at the national level.
They picked Bush I over Reagan in 1980. They picked Bob Dole and Pat Robertson both way ahead of Bush I in '88 -- when he had a record as the VP of the most liked President since Kennedy, and who'd won a landslide re-election in the previous cycle. They chose three candidates over the eventual nominee in '08, with hopeless Huckabee as their first choice. In '12, at least they almost called it, but still favored Santorum by a slim margin.
They did correctly predict Bush II in '00, Dole in '96, and Ford way back in '76. None of these correct predictions were running on a strong evangelical platform, however, the way that Carson is.
Turning to New Hampshire, their primary winner has gone on to secure the nomination in 5 of these 7 cases. They narrowly chose Buchanan over Dole in '96, and McCain over Bush II in '00. In neither case did they choose 2 or even 3 candidates over the eventual nominee, nor did their incorrect predictions choose an evangelical.
The most predictive of the early primaries, however, is South Carolina, where since 1980 every winner of the state's primary has won the nomination, except for Romney losing out to Newt Gingrich in '12, perhaps due to home-turf advantage in the Deep South.
Trump has enjoyed a healthy lead over Carson in New Hampshire, and an even wider double-digit lead in South Carolina. So even if Iowans do choose Carson over Trump, that won't matter in the big scheme of things.
By this point, Iowans are almost deliberately trying to choose someone who won't win, just to keep any one candidate from enjoying too much success. It's part of the extreme egalitarianism that the Scandinavians brought into the Upper Midwest (see the Law of Jante).