Saturday, June 7, 2008

ACCORD and AVANCE: Good News for Type 2 diabetes...really.

At least according to what's out there in the media and in the medical blogsphere on the two large recent diabetes trials presented at the ADA and published in the NEJM, in my opinion everyone seems to have missed the big picture, and the good (not bad) news about diabetes.

Just look at the headlines:

NY Times: Tight Rein on Blood Sugar Has No Heart Benefits
AP: Intense diabetes therapy didn't cut heart problems
MSNBC: Aggressive diabetes care doesn't prevent deaths
and my favorite
Forbes: Not So Sweet

Even Dr. Centor, in an excellent commentary on how the findings of these studies relate to problems using performance measures (which I agree with), does not mention anything positive..

The issue at hand is that though we use the hemoglobin A1c, a measure of overall sugar control, as a target for diabetes treatment which has been proven to prevent small vessel (microvascular) diabetes complications such as blindness and kidney problems, we have never proven (with scientific rigour) that it prevents larger vessel (macrovascular) disease such as heart attacks and stroke. The thought was that if there was a large study that treated even more aggressively (target A1c is usually recommend at less than 7%, normal A1c which is not currently a target is less than 6%), this might prove the point.

The Action to Control Cardiovascular Risk in Diabetes Study (ACCORD) looked at 10,251 type 2 diabetics (more than a third with a prior stroke or heart attack) with a median hemoglobin A1c of 8.1% and randomized them to receive intensive therapy (targeting an A1c below 6.0%) or standard therapy (targeting an A1c from 7.0 to 7.9%). At a year the groups achieved A1c's of 6.4% and 7.5% respectively, but after several years there was not real difference in combined hearts attacks, strokes or deaths in either groups (352 vs. 371), there was a statistical difference in death (257 vs 203), and the intense arm was stopped early.

The Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) trial was a similarly designed European study 11,140 patients with type 2 diabetes, except the targets for the intense group was less than 6.5% and not less then 6.0%. Interestingly the A1c's achieved by both sets of patients was nearly identical to ACCORD (6.5% vs.7.3%). The also found no difference in combined heat attacks, strokes and deaths; but fortunately found no difference in deaths from the intensive group as seen in ACCORD.

All of the reports (linked to above) point out that these studies failed to prove that aggressive treatment makes any difference in heart attacks or strokes, and in the case of ACCORD seemed to (surprisingly) cause more heart attacks. However, they are missing a majorly important finding of both studies. Treating diabetes aggressively is quite effective!

We already knew about microvascular disease, and ADVANCE showed us that this still holds to be true, with substantial reductions in early kidney disease. However, what is being overlooked is that both groups in both studies did much better than expected. In ACCORD, to show a statistical difference, the study enrolled enough patient based on research about event rates in diabetics. Based on their original calculations, there should have been about 500 heart attacks or strokes in the standard care group. However, there were only 371 in the standard care group, and even fewer in the intense arm. Death calculations were not published in the article, but I would imagine that though the death rate was (seeming paradoxically) higher in the intense arm, death rates in both arms were likely much lower than one would typically expect. Similarly, in ADVANCE, the found that after a mean of approximately 3 years of follow-up, it became apparent that the event rates (in the two groups combined) were lower than expected, and had to adjust their protocol, extending it by a year. Fewer events than anticipated in two very large studies seem like good news to me.

Why would these studies find lower event rates than expected? Likely because of more aggressive overall treatment. Not just lowering sugar, but lowering blood pressure and cholesterol levels with medications. In fact, because cholesterol and blood pressure reduction may be such important factors for macrovascular complications in diabetics, it might take a greater number of patients and a longer amount of time (not lowering the sugar to normal) to show the true impact, which is likely beneficial. Thus, patients with Type 2 diabetics should be relieved and not fearful of these reports. We have better medications that are more effective at lowering not only sugar, but also blood pressure and cholesterol and together (along with diet, exercise, and an aspirin) lower bad outcomes including heart attack and stroke, then we were previously able to accomplish just a few short years ago.

Why the difference in deaths?
There will be a lot of opinions about why patients in the intense arm of ACCORD had an increase number of deaths and ADVANCE did not. Here is mine.

The first two of the three the editorials which accompany both studies published in the NEJM note that most of the patients in the intense group were on a TZD like Avandia (the 1st and 3rd editorial both cite Dr. Nissen's well publicized/criticized meta-analysis about cardiovascular risk with Avandia-also published in NEJM), together suggesting that maybe the increased death rate was due to this class of medications. However, we know that TZD use can not account for the differences in outcomes in ADVANCE because as I have noted in a previous post on this study, when the study was stopped, given the concerns regarding Avandia, the NIH specifically looked for links to Avandia and found none. (It is interesting that this is not highlighted in article.) In fact, since most patients in the intense arm were taking a TZD like avandia, and heart attacks were lower, and increase in deaths could not be attributed to rosigliataozne, this should be most reassuring (see, more good news).

If not Avandia, then why the increased death rate in ACCORD but not ADVANCE. The second NEJM editorial points out that the rate of A1c drop was much more rapid in ACCORD then ADVANCE (1.4% within 4 months vs. 0.5% at 6 months respectively). How did they achieve such a rapid drop in one study, but not the other? This is reiterated in the discussion section of ADVANCE.

"Mechanisms speculated to underlie the excess mortality found with intensive glucose control in the ACCORD trial include the initial level of glycated hemoglobin, the degree and pace of glucose lowering, and the treatments used to achieve such lowering.20,22 In the ADVANCE trial, no subgroup of participants was identified to have evidence of an adverse effect of intensive glucose lowering on major vascular outcomes, including the subgroup with an initial median glycated hemoglobin value similar to that in the ACCORD study population.21"

Exactly how doctors tried to achieve normal A1c levels in Type 2 diabetics was not clarified in the ACCORD publication, an in fact, investigators could use their own clinical judgement. However in ADVANCE, this is spelled out clearly and in step wise fashion:

1. Add gliclazide (and SU like glyburide or glimeperide, but not used in the US)
2. Increase the dose of gliclazide MR (30-120 mg)
3. Add or increase the dose of metformin;
4. Add orincrease the dose of thiazolidinedione;
5. Add or increase the dose of alpha-glucosidase inhibitor;
6. Add bed time insulin therapy;
7. Add a full insulin regimen or increase the dose of a full insulin regimen

What you see hear is that in ADVANCE, adding insulin was the last resort in a stepwise fashion to get the A1c under 6.5%. Again, we don't know exactly what happened in ACCORD, but if you look at the differences in medications not only was there more TZD's in the intense group, there was a large difference in insulin use (77% in the intense group vs. 55% in the standard group). In ACCORD the rates of insulin use was 40% in the intense group and 24% in the intervention group. Hypoglycemia and weight gain (which are both caused by insulin) were reported as significantly increased adverse events in ACCORD.

Thus, I would argue that one of the reasons you might actually cause harm in intensively trying to lower sugar, is when you need insulin to get to that goal. This suggest (contrary to the ADA guidelines) that targeting to an A1c of less than 7.0% should be done with multiple medications, and not with insulin. In fact, it might be better to reserve insulin for those patients with an A1c of more than 8%, despite multiple oral agents (again, contrary to ADA guidelines).


Anonymous said...

You may want to look again at the evidence that A1C control in DM II "prevents blindness" or prevents "kidney problems."
Looking at the data suggests it does neither.

UKPDS: No clinically meaningful improvement in visual or renal function.


1% absolute difference in "nephropathy" (defined?) no creatinine differences or renal outcome differences. NNT 100 over 5 years to have one person have this likely meaningless outcome.

ACCORD: no microvascular issues addressed in study, as far as I know.

So the statement that "treating diabetes (type 2) aggressively is quite effective" is not supported by the best evidence we have.

Dr. Matthew Mintz said...

The statement prevents blindness and kidney problems refered to the UKPDS, and not ACCORD or ADVANCE, so I should have made this more clear.
Though retinopathy and microalbunuria are technically surrogates for blindness and renal failure respectively, they are biological surrogates. In otherword, we do know that untreated retinopathry will evenutally lead to blindness.
The BMJ article in your post makes several valid points about critical appraisal of information. Though the absolute risk reduction for microvascular disease is in the single digits for the entire study, this is not inconsequential (epectially is you are one of the 1 in 30 patients affected). In addition, the BMJ article points out that the NNT for obese patients treated with metformin for any outcome was 10, which is pretty substantial. In the US, most type diabetics are type 2 and most of them are obeses

Anonymous said...

You can maintain those surrogates possibly represent important clinical outcomes for patients in those studies. Others disagree. (as the links attest)

Bottom line, there were no clinically meaningful differences noted in vision or renal function. (and the retinopathy eval in UKPDS was not even blinded, btw) Focusing on subtle differences in surrogate markers isn't very convincing in this case, and certainly does not support the idea that tight A1C control is "important," whatever your opinion on the matter.

I agree with you on the value of metformin. That is independent of A1C, however, and a separate issue.

Dr. Matthew Mintz said...

I guess we will just disagree on how clinically meaningful retinopathy and microalbuminria are. I can tell you that after following patients for many years, those with early signs of these markers who get aggressive treatment do have fewer complications, and those that do not get appropriate treatment (because of non-compliance, lack of insurance, loss to follow up) are the ones that wind up in trouble.
There is no question that blood pressure control and cholesterol are critically important to macrovascular outcomes, and likely trump blood sugar control, even in diabetics.
Another important issue has to do with the time to progression of microvascular and macrovascular disease and progression of the disease itself.
These changes takes years to develop. In a newly diagnosed diabetic, it may take up to 10 years before you see microalbuminuria and retinopathy, and years after that to see kidney failure and blindness.
It you look at the UKPDS, patients in the "aggressive" arm starting losing control after the first year, and the majority of patients were not at goal after 3 years. Sulfonylureas and metformin do a good job at lowering blood sugar, but not so much when it comes to sustaining this over time. This is likely due to the fact that the underlying mechanism of type 2 diabetes is beta-cell dysfunction and insulin resistance, which are not addressed by either class of these drugs. The one class that does address this is the TZD's. However, when troglitazone was pulled from the market, this arm of UKPDS was stopped. Recently, the ADOPT study was published showing that rosiglitazone (Avandia) did a much better job at sustaining normoglycemia than SU or metformin.
Perhaps if we had long term (10 year) data on type 2 diabetics with sustained normal blood sugars, you would then be able to prove that tight glycemic control in type 2 diabetes prevents heart attack and strokes beyond the effects of statins and blood pressure medications.

Anonymous said...

well, ukpds showed a 3.2% absolute difference in outcomes, 2.7% of which was for non-blinded rates of retinal photocoagulation. No difference at all in visual acuity. 10 years. If this were indeed important, 10 years seems to be a decent amount of time to allow for some visual differences to develop. Additionally, the microalbumin improvement was minimal, most of the difference in surrogate outcomes was due to the photocoagulation.

combine this with ADVANCE, which showed no retinal differences and rather insignificant microalbumin differences, and it looks pretty bleak regarding the importance of tight control (with non-metformin meds) on patient oriented outcomes.

I appreciate what you have seen with your diabetic patients, but I suspect there are other risk factors in those patients you have "lost to follow up" over the years. Those who don't take care of their diabetes, skip appts, don't take meds, etc. probably do other things that aren't good for their diabetes. So I would expect worse outcomes for that group, as you have observed. That's why the data from randomized trials is important, to eliminate those sort of counfounders.

Christine said...

Dr. Mintz- I couldn't agree with your thoughts on this more.

Dr. Matthew Mintz said...

Apparently blogger will not let you post images into comments. However, if you go to the link below you can see some of the data from UKPDS. Patients started with normal A1c's. After 3 years many of the patients in the intense group had reverted back, and by 6 years, the average A1c was back to baseline. Thus, UKPDS shows that even for non-sustained glucose control for patients early in the disease, microvascular biologic markers were improved. This is impressive since the mean the A1c at the 10 year assessment was about 8%. Since microvascular complications take about 10 years, in order to show a real difference, you would need glycemic control for 10 years, which was not achieved in the UKPDS study.

Anonymous said...

Well, the difference in A1C between the conventional and aggressive groups persisted for 10 years, so maintaining a lower A1C didn't matter. Over 10 years. Which is the point.

You can always say "well, if they had been even more aggressive in the aggressive group, we would have seen differences." ACCORD said just that, and you know how that turned out.

Additionally, we have no reason to believe that the UKPDS differences became less important as the A1C rose in both groups. Conventional wisdom would suggest that oppsosite is true, i.e. you get more benefit moving from 9 to 8 on A1C than from 8 to 7.

So say it with me:
A sustained 10 year A1C difference in UKPDS didn't show any meaningful clinical differences in vision or kidney function.

Dr. Matthew Mintz said...

Though the A1c difference persisted in the 10 years, the intensively treated group was not at goal (A1c under 7) for most of the time. If they were to extend ADVANCE or ACCORD for another 6.5 years, then I bet you would see the "meaningful clinical differences" that you are looking for.
Also, retinopathy is not vision loss, but it is clinically meaningful. Even small changes in retinopathy in 4 years predict vision loss in the following 6 years (Arch Ophthalmol. 2001 Apr;119(4):576-8). Diabetic retinopathy is the leading cause of blindness in the US. I would say that this is pretty meaningful.

Anonymous said...

The "I bet you if they did the study differently they could show some differences", or "I bet if the study went longer you'd see some differences" doesn't hold much water. We can always say, "yeah, but" when there are negative studies. The evidence at this point (UKPDS, ACCORD, ADVANCE) is rather clear that we can't show tight A1C control with non-metformin meds in Type II DM does anything clinically important. These are well regarded studies in general and agreed to be the best we have regarding A1C control in DM II and outcomes. You are certainly entitled to believe tight A1C control is important, but the evidence isn't on your side and it then becomes your personal belief, not an evidence based opinion.

In addition, ADVANCE didn't show any meaningful retinal differences, and UKPDS wasn't even blinded when looking at retinal differences. So your persistence that tight control will ultimately save blindness in DM II is quite a leap. I would expect some difference in visual acuity over 10 years if this were indeed important.

Thanks for the discussion.