An estimated 655,000 Iraqis have died since 2003 who might still be alive but for the US-led invasion, according to a survey by a US university.
The research compares mortality rates before and after the invasion from 47 randomly chosen areas in Iraq.
The figure is considerably higher than estimates by official sources or the number of deaths reported in the media.
Critics have dismissed the findings because they are based on a statistical prediction rather than body counts.
Dr Gilbert Burnham of the John Hopkins Bloomberg School of Public Heath, based in Baltimore, says this method is more reliable, given the dangers of conducting thorough research in strife-torn Iraq.
The estimated death toll is equal to about 2.5% of Iraq's population, and averages out at more than 500 additional deaths a day since the start of the invasion.
Sharp rise
Researchers spoke to nearly 1,850 families, comprising more than 12,800 people in dozens of 40-household clusters around the country.
Of the 629 deaths they recorded among these families, 13% took place in the 14 months before the invasion and 87% in the 40 months afterwards.
Such a trend repeated nationwide would indicate a rise in annual death rates from 5.5 per 1,000 to 13.3 per 1,000.
The researchers say that in nearly 80% of the individual cases, family members produced death certificates to support their answers.
Reliable data is very hard to obtain in Iraq, where anti-US insurgents and sectarian death squads pose a grave danger to civilian researchers.
The survey updates earlier research using the same "cluster" technique which indicated that 100,000 Iraqis had died between the invasion and April 2004 - a figure that was also dismissed by many supporters of the US-led coalition.
'Survivor bias'
While critics point to the discrepancy between this and other independent surveys (such as Iraq Body Count's figure of 44-49,000 civilian deaths, based on media reports), the Bloomberg School team says its method may actually underestimate the true figure.
"Families, especially in households with combatants killed, could have hidden deaths. Under-reporting of infant deaths is a widespread concern in surveys of this type," the authors say.
"Entire households could have been killed, leading to survivor bias."
The survey suggests that most of the extra deaths - 601,000 - would have been the result of violence, mostly gunfire, and suggests that 31% could be attributable to action by US-led coalition forces.
The survey is to be published in a UK medical journal, the Lancet, on Thursday.
In an accompanying comment, the Lancet's Richard Horton acknowledges that the 2004 survey provoked controversy, but emphasises that the 2006 follow-up has been recommended by "four expert peers... with relatively minor revisions".
