February 24, 2008

How ADVANCE differs from ACCORD

As you all know by now, information about two major studies hit the press in the past few weeks. One, ACCORD, appeared to find that lowering the average A1c of a group of participants to 6.4% slightly increased the cardiovascular death rate.

The other study which lasted longer and enrolled twice as many people found that lowering the average A1c of study subjects to 6.4% resulted in no additional deaths.

There's been a lot of speculation about why these two studies came up with different results.

I did a bit of hunting to see if I could find out more about the details of these studies, and the little I have been able to find points to some important differences.

1. DIET. Contrary to speculation participants both studies were eating high carbohydrate diets. In ACCORD this was explicit. Nutritionists were assigned to hound study subjects on a regular schedule to eat a high carbohydrate/low fat diet.

In ADVANCE there was no formal dietary intervention. However, since the study was begun in 1999 at a time when it was a strongly held medical belief that low carb diets were dangerous for people with diabetes and since the study was conducted in countries where government-provided health care insisted patients follow a high carb/low fat diet, it is likely that participants in that study were eating a high carb diet, too.

2. BLOOD SUGAR DRUGS. According to the NIH patients in ACCORD were given "metformin; thiazolidinediones, or TZD's (rosiglitazone, pioglitazone); injectable insulins; sulfonylureas (gliclazide, glimepiride, glipizide, glyburide); and acarbose and exenatide."

Patients in ADVANCE were given only Gliclazide, (Diamicron), a sulfonylurea drug. Since it is dangerous to take this drug without a high carb diet, due to its ability to provoke serious hypos, this is more evidence that the diets in both studies were high carb/low fat.

But the fact that Avandia and Actos were not used in ADVANCE is suggestive.


Patients in ACCORD were given a wider selection of blood pressure drugs as the ACCORD design promoted using every possible drug available. I cannot find the explicit list of drugs, but given the design of the study, I would suspect it included ACE inhibitors, diuretics, and beta blockers.

Patients in ADVANCE took only a perindopril-indapamide combination. The first is an ACE inhibitor, the second a diuretic.


Patients in ACCORD were put on high doses of statins. Then a fibrate was added to this.

Patients in ADVANCE were NOT given any cholesterol treatment as part of the study.


The ACCORD study violated all the principles of good science because it introduced far too many independent variables. ADVANCE limited the variables under study making it much easier to interpret their effect.

The subjects in ACCORD were taking Avandia, Actos, and Byetta all of which are new. We now know that Avandia increases heart attack. We also know that Actos and Avandia both cause heart failure in people who didn't have it before starting the drug. Byetta is new enough that we don't know what it's impact on the heart is, long term.

If ACCORD did use beta blockers, that may be a factor in the increased deaths too. While these drugs on their own can prevent heart attack, reading their Prescribing Information makes it clear that they can also trigger heart attack in some situations. Also, they can be toxic to people whose livers aren't working properly. Avandia, Actos and statins can all affect liver function. When all these drugs are taken together, it is possible that small changes in liver function might have had an impact on the effects of beta blockers. (Or, for that matter, any other drug being taken.)

Finally it is possible that the use of STATINS and fibrates to pursue aggressive cholesterol control in the ACCORD trial is at fault. Or that these drugs work synergistically with some of the blood sugar drugs to cause problems.

The Vytorin study found slightly more growth of plaque in the arteries of people pursuing more aggressive cholesterol control.

But the study design of ACCORD is so sloppy, it is not likely we will ever know exactly what the problem with its protocol was.

For details on the study protocols visit:

ACCORD: NIH: ACCORD Questions & Answers

ADVANCE Study Rationale and Design of ADVANCE


Anonymous said...

The fact that ACCORD used TZDs is suggestive. They were aware of that and claimed that they analyzed for that and found no effect.

As for the other drugs, not everyone got both BG-lowering drugs and cholesterol-lowering drugs or BP-lowering drugs.

You can find the details of the study protocol at the ACCORD Web site. But I don't think we'll know what factors were important until we can also see the analysis of the results of each of the groups.

Some got standard lipid treatment and standard BG treatment. Others got standard lipid treatment and intensive BG treatment. Some got intensive lipid treatment and standard BG treatment. Others got intensive lipid treatment and intensive BG treatment.

No one got intensive BG treatment and intensive BP treatment and intensive lipid-lowering treatment.

Jenny said...


No amount of statistical analysis can overcome the multiplicity of interacting variables introduced into ACCORD.

The study was designed in such a way that it could only be valuable if the scientists got the result they assumed they would get. Then they could have concluded "Lowering A1c prevents heart attack."

Red Sphynx said...


Thanks so much for the effort you've put into analyzing this. Nobody but nobody else – in the journals, in the science-writing community, in the major newspapers, in the blogs – is bringing in the combination of knowledge and hard work that you've turned to this issue. Thanks. A lot.

Adam Becker Sr