Primary Care Corner with Geoffrey Modest MD: Empaglifozin, the good and the bad

By Dr. Geoffrey Modest

A recent study looked at the SGLT2 inhibitor (sodium-glucose cotransporter 2) empagliflozin and cardiovascular outcomes/mortality in patients with type 2 diabetes (see N Engl J Med 2015;373:2117​). The SGLTs are involved in active renal transport of glucose in the kidneys, and SGLT2 is the most important one (reabsorbs 90% of the glucose filtered by the glomeruli) and seems to be pretty kidney-specific (SGLT1 is less important in active glucose transport, getting the remaining 10%, and is more widely expressed and therefore more likely to have broader physiologic effects and adverse effects if blocked). The context of this study is that there are a slew of studies showing markedly increased cardiovascular (CV) mortality in patients with type 2 diabetes, this relationship seems to be dose-related, and the increase extends to those with pre-diabetes/glucose intolerance. Other studies have shown that empagliflozin is associated with weight loss, reductions in blood pressure, increases in both LDL and HDL, and decreases in arterial stiffness/vascular resistance, visceral adiposity, albuminuria, and plasma urate levels (all of these might contribute to CV disease).

Details of the current study:

  • 7020 patients (mean age 62, 72% white/22% asian/18% latino, BMI 30.7, A1c 8.08%, 74% on metformin/49% insulin with mean dose 52 units/43% sulfonylurea/11% DPP-4 inhibitor/4% TZD, 95% hypertensive, 80% on lipid-lowering therapy. 83% on aspirin), all with established cardiovascular disease, all with A1C=7-9 and not on meds for >12 weeks or A1C=7-10% and on stable diabetes meds for >12 weeks). Subjects were randomized to empagliflozin 10mg, 25mg, or placebo daily, followed 3.1 years
  • Excluded: BMI >45, eGFR <30,  fasting glucose >240, ALT >3x upper limit of normal


  • 4% prematurely stopped the study drug (similar to placebo)
  • Weight decreased with med from 86 to 84 kg, waist circumference from 105 to 103 cm, SBP from 135 to 132 mmHg, DBP from 76 to 75 mmHg, LDL increased from 88 to 90 mg/dL and HDL from 44 to 46 mg/dL, uric acid decreased from 6 to 5.7 mg/dL
  • In the group on placebo, 31.5% had changes in their other diabetic meds, esp. sulfonylureas (increased 3.8% in those on empagliflozin and 7.0% on placebo), insulin (11.5% vs 5.8%), TZD (not sure which used, but increased 2.9% vs 1.2%), DPP-4 increased (8.3 vs 5.6%)
  • A1c was about 0.5% lower in those on empagliflozin, though there were changes as noted in other diabetic meds given
  • ​Primary outcome (composite of death from CV causes, nonfatal MI or nonfatal stroke) in 490 of 4687 (10.5%) of those in the pooled empagliflozin groups, vs 282 of 2333 patients on placebo (12.1%), so HR 0.86 (0.74-0.99; p=0.04)
  • No difference in rates of MI or stroke, but a lower death rate from CV causes (3.7% vs 5.9%, relative risk reduction RRR of 38%), hospitalization for heart failure (2.7% vs 4.1%, RRR of 35%), and death from any cause (5.7% vs 8.3%, RRR of 32%)
  • No difference in secondary outcome (primary outcome plus hospitalization for unstable angina)
  • Adverse events: increase in genital infections (5% in men vs 1.5% on placebo, 10% in women vs 2.6% on placebo). Rest nonsignificant.

So, a few issues (many of which were buried in the supplemental materials, which seems to be happening a lot in the past few years and my guess is that relatively few readers delve into these details, in part because it is a hassle, in part because it requires full, and expensive, access to the journals, and also is on the internet only. BUT it is often hard to really interpret the studies without seeing them…)

  • Of the very significant difference in CV deaths (5.9% in placebo and 3.7% with empagliflozin), the largest single category was “other cardiovascular death” (2.4% vs 1.6%) which “includes fatal cases that were not assessable due to a lack of information and were presumed to be cardiovascular deaths as per conventional definition”. And much smaller numbers of other CV events (acute MI in 0.5 vs 0.3%, worsening in heart failure in 0.8 vs 0.2%, stroke in 0.5 vs 0.3%). these increases in “presumed” events really undercuts the significance of the cardiac effect, especially in light of the very small numbers of events in these very high risk patients over >3 years of follow-up
  • The beneficial effects of the empagliflozin were within 3-6 months overall, a very rapid change. This makes the change in A1C from increased excretion of glucose in the urine an unlikely candidate for the CV benefit, and perhaps a quick change in arterial stiffness is more likely, or a decrease in arrhythmias for unclear reasons (sudden death was the most common single cardiac death event, with 1.6% in those on placebo vs 1.3% in those on empagliflozin). But one might imagine an even more profound cardiovascular benefit by having more of the patients on statins (only 76% of this really high risk population were on statins!!!)
  • Unclear what thiazolidinedione was used. Rosiglitazone??? (Which has more CV events). Subgroup analysis did find a nonsignificant trend to increased CV death in those on TZDs overall, though the interpretation really depends on which is being prescribed (pioglitazone may actually decrease cardiovasc events, see next point)
  • I think the study was not really accurately framed. their comment that evidence of lowering blood sugar “reduces the rate of cardiovascular events and death has not been convincingly shown” in fact does have some reasonable support for a few agents:
    • Metformin in UKPDS: there was a 16% reduction in cardiovascular complications — combined fatal or nonfatal MI and sudden death with metformin (p=0.052), was borderline significant. In a further analysis, patients allocated to metformin, compared with the conventional group (including insulin and sulfonylurea), had risk reductions of 42% for diabetes-related death (p=0·017), and 36% for all-cause mortality (p=0·011) (seeLancet 1998; 352: 854)
    • A large VA study compared monotherapy with sulfonylureas and metformin, finding a 21% increased hazard ratio for CV events or deaths in those on sulfonylureas vs those on metformin (see Ann Intern Med. 2012;157:601-610.)
    • Pioglitazone: in the PROACTIVE study, the main secondary endpoint was the composite of all-cause mortality, non-fatal myocardial infarction, and stroke. 301 patients in the pioglitazone group and 358 in the placebo group reached this endpoint (0·84, p=0·027) (see Lancet 2005; 366: 1279)
  • Not to surprise you too much, but the cost of this drug is actually a lot: $5000/year with very low absolute benefit
  • And, within days of release of this article, the FDA released a warning for SGLT2 inhibitors (see, noting the risks for ketoacidosis and serious urinary tract infections. Updating their May 2015 warning, the FDA noted that 73 cases of diabetic ketoacidosis and 19 cases of urosepsis and pyelonephritis that started as UTIs have now been reported in patients taking these drugs, with some needing ICU treatment and dialysis (and this undoubtedly does not include all cases). So, be aware of these issues and stop the med/treat promptly​. They are now requiring postmarketing safety studies for 5 years.  But, as noted in prior blog , <20% of required postmarketing studies for medical devices were done within 3 years of the FDA requiring them, and no fines have been issued for noncompliance with the FDA mandates.  Also, as a point of reference regarding the DKA issue: noted the increase in ketoacidosis with SGLT2 inhibitors, with 1/2 the cases without any typical DKA triggering factors (e.g. infection), events happened in some with only very mildly elevated blood sugars (including <200 mg/dL), and ketoacidosis ranged  from 1-175 days after starting the med.

So, what is one to do??? As many of you may be aware, I am pretty hesitant to jump on the bandwagon of new drugs, especially when we have drugs that work pretty well. There are a lot of questions above about this study, ranging from what additional drugs were used in the placebo group (more sulfonylureas, which don’t seem to have much CV benefit and may even have some harm; and ??rosiglitazone — unclear if this was used). Also the likely mechanism of action was not by lowering the A1c, given the rapidity of decreasing cardiovascular events, so is this the best drug to achieve the benefit? Should more people be on statins? And SGLT2 drugs are not necessarily benign (hence the FDA warning).​

(Visited 7 times, 1 visits today)