Primary Care Corner with Geoffrey Modest MD: Hemoglobin A1c reduction and decreased cardiovascular event

​by Dr Geoffrey Modest

A recent cohort study found an association between those patients who achieved an early reduction in HbA1c with metformin and a significant reduction in subsequent cardiovascular events (see DOI: 10.2337/dc16-2271).



— a population-based cohort study in northern Denmark from 2000-2012 assessed 24,752 new-onset diabetics 6 months after initiating metformin therapy

— median age 62.5, 55% males. Median follow-up 2.6 years

— primary endpoint: subsequent rates of acute myocardial infarction, stroke, or death, controlling for baseline HbA1c and other confounding factors (age, sex, year of starting metformin therapy, microvascular and macrovascular complications, obesity, alcoholism, antiplatelet drugs, statins, antihypertensives, psychiatric medications, achieved cholesterol target, and use of other unspecified glucose lowering therapy). socioeconomic status was adjusted by using education as a proxy.



— those who achieved A1c <6.5% tended to be older (> 70yo), female, and more likely to have initiated metformin later in the study (between 2010-2012), tended to have slightly more microvascular and macrovascular complications at baseline, take more preventative medications, and have less obesity. They also tended to have lower baseline A1c initially.

— those with the greatest point reductions in A1c tended to be younger, had lower prevalence of macrovascular complications, less comorbidity, and took fewer preventive medications. They also tended to have the highest baseline A1c and received more additional glucose lowering therapy.

— During the follow-up, there were 439 incident MIs, 594 strokes, and 1845 deaths.

— the risk of a primary end-point, compared with an achieved HbA1c <6.5% in the adjusted model, was:

— 6.5-7%: 18% increased risk, HR 1.18 (1.07-1.30)

— 7-7. 5%: 23% increased risk, HR 1.23 (1.09-1.40)

— 7.5-8%: 34% increased risk, HR 1.34 (1.14-1.57)

— >8%: 59% increased risk, HR 1.59 (1.37-1.84)

— these results were consistent for the individual outcomes, age groups, presence or absence of comorbidity at baseline, or socioeconomic status.

— the clearest association between higher A1c and worse clinical outcomes was in patients >70 years old.

— a 4% absolute decrease in HbA1c was associated with a 20% decreased risk of a primary outcome; lesser degrees of risk reduction were not statistically significant



— observational data from many studies are quite consistent that there is an increased risk of atherosclerotic disease in diabetics, and that this increase starts early in the prediabeticrange (eg A1c 5.5-6), especially in men, and increases with increasing A1c. The data on decreasing risk by lowering A1c is less clear, and seems to be dependent on the medication used. Some medications have some reasonable support for cardiovascular benefit, including metformin, pioglitazone, and GLP-1 agonists (I am very suspect about empaglifozin, see below). It is important to know that the published target of getting the A1c’s as low as possible, around 7% or even lower, are based on decreasing risk of microvascular complications found in several studies, not the really important macrovascular ones (80% of diabetics die from cardiovascular causes). Which is not to say that microvascular complications are not important…

–As noted in prior blogs, the ACCORD study questioned the utility of lower A1c targets for macrovascular protection but (I think) is overquoted and has led to a more lackadaisical approach to diabetes management. See blog noting that “A subgroup analysis of this ACCORD study actually found that those who achieved a lower A1c in fact did better (in terms of cardiovascular endpoints), all the way down to an A1c of 6!!, but as the number of meds needed in the attempt to lower the A1c increased, they had worse outcomes (i.e., those in the intensive therapy group who had medication-flogging to improve their A1c had worse outcomes even at a much higher A1c). See Riddle MC. Diabetes Care; 33:983.” Also see the blog which goes into more detail on ACCORD. So, my conclusion was that the lower the A1c, the better in terms of macrovascular complications in ACCORD, but only if this lower A1c can be achieved easily with meds (and, reinforcing the above comment, the choice of meds may be extremely important. For example, in the ACCORD study, 91% of those in the aggressive treatment arm took TZDs, and mostly rosiglitazone, known to increase cardiac mortality!!!. Interestingly in this current study, those >70yo did better cardiovascular-wise if their A1c were low with metformin, which suggests that age may not be the decisive factor in targeting A1c goals, but perhaps other things (comorbidities, longevity predictions), though my sense is that currently clinicians are somewhat more hesitant to be aggressive with older patients, quoting the ACCORD trial

–one peculiar aspect of this study to me was why did so many patients have high A1c’s 6 months after starting metformin? My clinical experience, with a few dozen new-onset diabetics (though not the 24.7K in this study), is that they uniformly respond to metformin with essentially normalization of their A1c, independent of how high their initial A1c was. Even patients with A1c’s in the 15-18% range decrease to <7% with just metformin 500 mg once a day. The purported explanation for this is that they have glucotoxicity, whereby the increased blood sugar leads to decreased insulin effectiveness (as shown, for example, in decreased insulin-mediated glucose uptake into muscle), and this is true in laboratory situations with both diabetics and nondiabetics.  Typically new-onset diabetics need insulin injections early on to help decrease the blood sugar, thereby decreasing the glucotoxicity for which their struggling pancreas cannot compensate with more insulin production, and then, when their blood sugars are lower, their marginal endogenous insulin reserves work. And in my experience exogenous insulin can uniformly be stopped, typically within 1-2 months, and some do not even need to continue the metformin (at least for months to years, when their pancreases get increasingly tired).

–and, why did those with high A1c’s 6 months after starting metformin do so poorly? was it simply poor glucose control? or, were those who did not achieve a low A1c fundamentally different from those who did, perhaps in unmeasured ways?? did they have diabetes longer, leading to more atherosclerosis (ie, was the increase in cardiovascular events really just lead-time bias, where those with longer duration of diabetes before diagnosis had more atherosclerotic disease just because they had diabetes for longer)? Or, is their diabetes fundamentally different: do the ones who easily achieve really low A1c levels have a different, “mild” type of diabetes and would not get heart disease for a really long time anyway, while others have great difficulty getting their diabetes controlled, have more complications, but the complications have nothing to do with the A1c level achieved but with the fact that they just have “aggressive” diabetes???

–also, unfortunately, in the above study there were obvious important unmeasured factors, including smoking, BMI, diet, physical activity, social support, patient motivation, etc. which really could affect their results. And there may also have been an inherent bias that clinicians tried to get better blood sugar control in patients who were healthier and they felt had a better prognosis, so actually had less cardiovascular disease because of that? (this is hard to sort out in an observational study as this one)

—  As a side issue, studies like this one from Denmark, as well as others from most European countries, showcase the advantages of their having extensive medical databases, linkages to prescription data, linkages to social/demographic data, and nearly 100% followup.  The US, with the most expensive medical system in the world, unfortunately pales in comparison (?reflecting different priorities in public health???)…


so, why do I bring up this article, given the above limits of the interpretability of such an observational study?  a few reasons:

–I think there probably is value to more aggressive diabetes treatment to lowering cardiovascular events (and, again, this is what kills diabetic patients). and I think the results of the ACCORD trial weigh much too heavily on the various guidelines, since this study (i think) was quite flawed

–I think that we should really move towards using diabetes medications which have the clearest cardiovascular benefit (eg metformin, GLP-1 agonists, as opposed to insulin and sulfonylureas) as a priority, vs just focusing on achieving the best “number” for our A1c’s (untested, just my thought), though it is clear that A1c level is important for the microvascular complications

–I think studies like this one should stimulate us to question our model of disease (diabetes, here). Is it one disease or a combination of different ones with different outcomes but we lump them together (sort of like breast cancer: some are really aggressive and spread even before we can even detect them. some slowly increase in size and are easily treated but may not change life expectancy much…..)?

–and, I think it is a clinically useful and important intellectual activity to look at studies like this one (which I do think has some validity) in order to reflect on and critique the study; think about its potential biases, assumptions and limitations to its generalizability; and figure out if or how it should be incorporated into our understanding of the disease and how we should treat patients (though oftentimes these studies raise more questions than they answer, but maybe the questions are more advanced as we see more studies and get more data…..).  This is one of my main impeti (?plural for impetus) for doing these blogs. I think that overall in medicine we are deluged/overwhelmed daily with huge numbers of medical articles, guidelines, etc. And my concerns are that in many cases we just hear about/read the bottom line: does the med work?? etc. And the studies are mostly funded by drug companies. And they design them to optimally get the result they want. And many of the details of the study are buried in the text (and, more and more, buried in the supplemental material, for which one needs either a subscription to the journal or access to a medical library, but in any event takes even longer to sort out). See empagliflozin blog listed below as one of many examples of what I think is a poorly designed drug-company sponsored study. And nowadays the drug companies seem to have greater access to the media, who also publish a not-so-in-depth review, and they certainly have much more access to direct-to-consumer advertising, both of which may push us to prescribe these drugs.  So, all in all, it is getting much harder to read the onslaught of medical information critically, decipher it, and apply it appropriately. And I hope these blogs help a little …..


See here and here for blogs critiquing the empaglifozin study, suggesting that the accepted cardiovascular benefit is not so clear when one scratches the surface of the article. It also comments on the benefits found with metformin and pioglitazone in other studies.  Though I should add that the FDA does have warnings about these SGLT-2 inhibitors (such as empagliflozin) causing acute renal injury, ketoacidosis, urosepsis/pyelonephritis, and genital mycotic infections  (see


See blog on liraglutide, a GLP-1 agonist, which I think really does have significant cardiovascular benefit, as well as powerfully lowering A1c levels

(Visited 4 times, 1 visits today)