Home / Resources / Articles / Culprit Identified in ACCORD Study

Culprit Identified in ACCORD Study

May 13, 2010

It was not hypoglycemia from intensive control or intensive control itself that caused the increased deaths in the ACCORD study. This information will not be reported on every TV station as the original study results were, so it is your responsibility to inform your patients that intensive control that is individualized by your diabetes team is safe…. 

First author of the new paper, Dr. Matthew C. Riddle commented that, “Newly published findings show that, rather than hypoglycemia being the culprit, as was originally thought…. This seems to be a problem that was specific to the subgroup of people who tried the intensive strategy but could not get their HbA1c levels down below 7, those who had persistently high HbA1c levels.” So it wasn’t really intensive therapy that was associated with the increased mortality, it was unsuccessful intensive therapy, he said.


ACCORD Glycemia was conducted in 10,000 high-risk patients with Type 2 diabetes who were randomly assigned to either intensive glycemic control (HbA1c <6%) or standard glycemic control (HbA1c 7%-7.9%). To everyone’s surprise, it was stopped early, in February 2008, because of higher mortality in the intensive-glycemic-control group.

First author of the new paper, Dr. Matthew C. Riddle (Oregon Health and Science University, Portland), says, “There was a mismatch between the better level of glucose control achieved in the intensive arm, the epidemiologic evidence that in general lower is better, and the fact that some people in the intensive arm had bad outcomes — it seemed counterintuitive.”

The risk of death with the intensive strategy increased approximately linearly from an HbA1c of 6% to 9%, the new paper shows. And mortality was greater in the intensive group, compared with the standard arm, only when average HbA1c was >7%. Those who easily achieved target A1c levels had the lowest risk of death, while those who struggled to achieve low A1c levels had the highest risk of death, Riddle said.

In an editorial accompanying the new paper], Dr. Edward J. Boyko (VA Puget Sound Health Care System, Seattle) agrees that the study likely exonerates hypoglycemia as the cause of the excess number of deaths in the intensive-treatment arm of ACCORD glycemia.

And American Diabetes Association spokesperson Dr. Richard Bergenstal (International Diabetes Center, Minneapolis, MN), who was involved with the ACCORD Glycemia trial but not with this new analysis, stated that, “When ACCORD was reported, everybody said, ‘Oh, I’m sure it was the rapid drop in the A1c,’ ‘I’m sure it was the lower A1c,’ ‘I’m sure it was the hypoglycemia,’ and none of those has proven to be true. It was important to see these data, to say that in the intensive group it really was not the people with lower A1c who had problems, it actually was those who had the higher A1c who, despite intense efforts, we couldn’t get under control.”

“That doesn’t mean we don’t still have unanswered questions,” Bergenstal adds, “but at least we are starting to address some of the hypotheses that were generated when that study finding came out.”

“The primary end point was the relationship between the treatment strategy — intensive or standard glycemic control — and a cardiovascular composite outcome, which included cardiovascular death,” says Riddle. “It looks like in the three-and-a-half-year follow-up and also in the five-year analysis done more recently, that there is a trend toward benefit with the CV composite outcome, and yet there was the short-term outcome of increased mortality in the intensive arm,” he explained.

“We have thought of a composite reflecting all the different components, but in this case, the composite and the short-term mortality went in different directions, and that’s troubling both clinically and to the trialist.

“I hope we will be able to make some sense of this dilemma; I have come to believe there is a lot in the excellent ACCORD data set, and we need to keep working to understand what it means.”

Riddle says the findings so far, along with recent data from the ACCORD BP and ACCORD Lipidparts of the trial, can be taken together to indicate that this population, consisting of people who have had diabetes for 10 years with established medical complications, “is difficult to help in a short-term kind of way. You don’t reverse what’s already happened, and that’s an important message we can use, practically speaking.”

“When the glucose part of the study was stopped early due to excess mortality, this deflated a lot of people, because we thought, ‘Oh, this is going in the wrong direction, it’s disappointing, but perhaps the BP and Lipid arms will show more positive findings.’ But, in fact, neither of these arms — both reported recently at the ACC 2010 meeting — showed any benefit to be gained from either intensively lowering systolic blood pressure to less than 120 mm Hg or adding fenofibrate to statin therapy in this patient population.”

“This gives a stronger momentum to the idea that we need to be thinking that one size doesn’t fit all, we need to have different targets for different groups of people and perhaps different treatment strategies to reach those different targets as well.”

“This means we don’t have to be fearful about treating early in the natural history of diabetes and keeping HbA1c low — if we could assume people with newly recognized diabetes can realistically keep the HbA1c in the range between 6% and 7% — but those with a 10-year duration of diabetes should aim for a higher target in general, which is helpful,” Riddle says.

“This is something of a new idea, because previously there has been a strong impetus to having standardized guidelines for doctors and people with diabetes, but it’s probably not the right thing to do,” he adds.

Bergenstal agrees to an extent, “For patients who’ve had diabetes for a number of years and are at somewhat high risk for CV disease, we should not be aiming for a normal HbA1c of <6%, so that is an important lesson, number one. But that doesn't mean that because we shouldn't be aiming for <6%, that we shouldn't be aiming for good control, and I still think we are settled around 7%, because the data indicate that prevents microvascular disease."

“I think we should be individualizing therapy, but from a starting point,” he continues. “And for the general population, right around 7% is a good goal. If we can get to 7% without excessive therapy, then I think we should. But if somebody is above 7% or 8% and you are pushing very hard, then at some point you say, ‘In this person, we need to back off a bit and maybe their goal shouldn’t be less than 7%.’ With those caveats, we have some better practice guidelines than we did the minute the study first came out and everybody said, ‘I guess glucose isn’t important.’ “

However, Riddle says, “There still remains the problem of how to identify those people who are at higher risk whom we shouldn’t be so aggressive with. We don’t have the answer to that yet, but I think the motivation is there for us now to go back and keep doing analyses to try and understand the problem.”

Bergenstal concurs, “You have to ask some very pointed questions, and then you can start to identify those at higher risk.”

Diabetes Care May 2010; 33:983-990: Boyko EJ. ACCORD glycemia results continue to puzzle. Diabetes Care 33;1149-1150.