intensive glucose control and heart disease

the Lancet recently had a posthoc analysis of the ACCORD trial, finding that decreasing A1C in type 2 diabetics was associated with decreasing ischemic heart disease events (see dm ACCORD intensive control dec CAD lancet 2014 in dropbox, or Lancet 2014; 384: 1936–41). This was one of 3 contemporary trials of intensive vs standard control of diabetics -- neither ADVANCE nor VADT studies showing benefit of intensive diabetic control for macrovascular complications, and ACCORD actually found an increase in mortality in the intensive group. 

The trial:
--10251 patients, mean age 62, with diabetes on average of 10 years, and all with high risk for heart disease and 35% with known CAD (coronary artery disease), and baseline A1C of 8.1% (this trial also had blood pressure and lipid trials embedded). no difference in # on statins (88%) or aspirin (76%).
--intervention: keep adding meds to achieve an A1C of 6.0% in intensive group vs 7-7.9% in standard group
--the achieved A1C was substantially different in the intensive group (6.4%) vs standard (7.5%)
--to achieve this difference: 77% vs 55% on insulin, 91% vs 58% on thiazolidinediones (TZDs). although not highlighted in this article or most of the others i've seen on the ACCORD study, rosiglitazone was almost always the TZD of choice!!.
--the trial was terminated early, after 3.5 years, because of a 22% increase in mortality overall

Post-hoc analysis results:  the 5-year incidence of
--ischemic heart disease 13% lower in intensive group
--myocardial infaction 16% lower in intensive group
--coronary revascularization 16% lower in intensive group
--unstable angina 19% lower in intensive group
--and, all of the above curves were diverging over time (ie, intensive group doing increasingly better)
--no significant effect on fatal MI or new angina

so, how does one put this all together and apply it to primary care patients?
--it should be stressed that any posthoc analysis is potentially flawed by unsuspected biases
--it is pretty clear from many observational studies that there is a dose-response curve, with higher A1C associated with more heart disease. it is important to recognize that those with higher A1C may also have more comorbidities, may not take care of themselves in many different ways which may significantly adversely affects their health, etc (ie, these observational studies may have an inherent bias, thereby overstating the A1c/CAD relationship). but in this regard, it is interesting that there have been a few studies finding that even mildly elevated A1Cs, even in the prediabetic range, are associated with more heart disease. In a UK observational study for example, comparing those with A1C <5.0, men with A1C of 5.5-5.9 had a 56% increased risk of CAD events and overall mortality, those with A1C of 6-6.4 had 80% increase, and so on. fewer women had events in general, and had a similar trend with A1C but did not reach significance til A1C >7 (see dm A1C and cardiovasc dz annals 2004 in dropbox, or Ann Intern Med. 2004;141:413-420). these studies were also observational and potentially biased, though i do treat a patient with an A1C in the high five to six range more aggressively in terms of pushing for lifestyle changes and even therapy (eg i am more likely to prescribe a statin in someone with highish A1C and otherwise marginal indications than if the A1C were lower).
--another interesting posthoc analysis of the ACCORD trial also found that lower achieved A1C was associated with fewer cardiovascular events.  This study found that the increased mortality differences found in ACCORD between the standard and intensive groups could be explained by which group the patients were assigned to!!!  that is, although the achieved A1C did matter in terms of cardiac outcomes (lower A1C, fewer outcomes), those flogged with more meds to achieve it did less well, and those who easily achieved it did better (see Diabetes Care 2010; 33: 2722-24)​.
--so, how does one reconcile the overall study conclusion that intensive therapy in the ACCORD trial is associated with a 22% increased mortality, yet in the posthoc analysis the specific items above were better with intensive therapy? there was an interesting Italian study in 2009 (see dm italian study comorbid a1c goal annals 2009​ in dropbox, or Ann Intern Med. 2009;151:854-860) which found that for those diabetics with few medical comorbidities, those with A1C <6.5 had fewer cardiac events. in those with many comorbidities, there was no difference if A1C <6.5, or < 7.5 or higher (ie, a lower A1C did not matter). so, it is possible that the group on intensive therapy in ACCORD included 2 subgroups: one with very difficult-to-control diabetes who perhaps had worse outcomes by flogging them with more meds and one easily controlled who actually did better, but that the totals for the group was overall a poorer outcome. there might also have been a role played by rosiglitazone, which seems to cause heart disease (and, i think, be taken completely off the market). in addition, those in the intensive group with more difficult-to-treat diabetes may be  different physiologically (eg, they had more comorbidities and perhaps more systemic pathologic effects of the diabetes) or behaviorally (less adherent to meds, lifestyle, etc) than those with easier-to-treat diabetes.
--my fear is that the negative publicity about the ACCORD trial (especially in conjunction with the ADVANCE and VADT trials) has led to more complacency about pushing for lower A1C goals.  my sense is that if lower A1C's can be achieved easily and with better drugs (especially metformin, then moving to sulfonylureas/insulin, or even pioglitazone as in the PROACTIVE study), and especially in younger/healthier individuals, then that is very likely to be good not just for preventing microvascular complications (which are pretty clearly better with lower A1C levels) but also with heart disease. and that adding on more and more medications (flogging) just to decrease the A1C may in fact be harmful. A1C is a diabetes marker, but our goal as providers is to treat patients, not markers. it seems to me that the above analysis highlights that we should be treating the whole patient, taking into account their age, comorbidities, lifestyle, etc and not just adding on more and more medicines just to get the A1C lower (though, unfortunately, measuring A1C is the low-hanging fruit that is so easy for outside organizations to rate us on).
--of note, there is also an article in the current Evidence-Based Medicine journal by Francesco Giogino (see 10.1136/ebmed-2014-110010) finding that the increased cardiovascular mortality in the ACCORD trial was predominantly in younger and not older individuals, though the risk of hypoglycemia was higher in the older group.