Pool every randomized trial of liraglutide in heart surgery and the drug moved nothing patients feel. It did not cut deaths, complications, cardiac events, or even the low blood sugar it was picked to avoid.
That is the finding of a meta-analysis published September 4 in Cardiology in Review ↗. The authors gathered four randomized trials of subcutaneous liraglutide ↗, a GLP-1 drug better known as Victoza and Saxenda, given around cardiac surgery, and pooled 446 patients. On every hard outcome the drug and the comparison arm came out the same.
Why anyone tried this
High blood sugar during and after heart surgery is dangerous in a concrete way. It tracks with wound infections, acute kidney injury, and death. The standard fix is intravenous insulin, but insulin carries its own hazard: push glucose down too far and you get hypoglycemia, a low that can itself harm a fragile post-surgical patient.
GLP-1 drugs offered an elegant escape. They tell the pancreas to release insulin only when glucose is elevated, and back off when it is normal. In theory that separates the thing you want (glucose control) from the thing you fear (a dangerous low). A drug that steers glucose without overshooting is exactly what a cardiac operating room wants.
The mechanism does what it says on the glucose meter. In the largest of the pooled trials, the Dutch GLOBE trial ↗, preoperative liraglutide cut the share of patients who needed extra insulin from 61 percent to 43 percent while holding glucose in range. Better numbers, fewer insulin drips.
Where better numbers stopped
The meta-analysis asked the next question: did those better glucose numbers turn into better outcomes? Across the four trials, no.
Thirty-day mortality was 1 death in 161 liraglutide patients versus 3 in 160 controls, a risk ratio of 0.42 but with a confidence interval running from 0.06 to 2.81. The composite of any postoperative complication was essentially flat (risk ratio 0.92, from 0.74 to 1.14). Cardiac adverse events did not move (1.08). Hypoglycemia, the specific harm the drug's mechanism was supposed to sidestep, did not move either: 8 events in 165 patients versus 9 in 166, a risk ratio of 0.85 that ranged from 0.34 to 2.13. Nausea and vomiting produced a risk ratio of 3.01 with a confidence interval stretching to 35, which is a statistician's way of saying the trials saw too few events to know anything.
Read those intervals and the real result comes into focus. This is not a drug that failed. It is an evidence base too small to answer the question. A risk ratio of 0.42 for death sounds like a large benefit, but a range that spans from a 94 percent reduction to a near-tripling of risk contains every possible conclusion. With 446 patients and a handful of events, the pool cannot detect anything short of an enormous effect, and no cardiac-surgery outcome produces enormous effects.
The pattern worth naming
Observational data keeps hinting that GLP-1 drugs help surgical patients. A database study found that patients on GLP-1 before carotid artery stenting had fewer deaths ↗, though the benefit turned out to be all-cause mortality rather than the strokes and heart attacks the procedure targets, the fingerprint of healthier patients getting the drug. Randomized trials are supposed to strip that bias out. Here the randomized evidence, the kind that should settle the question, is simply too thin to say either way.
That is the useful takeaway for anyone weighing GLP-1 drugs as perioperative glucose control. The mechanism is sound and the glucose numbers improve. Whether that prevents a single infection, kidney injury, or death remains unproven, not because a trial showed it does not, but because no one has run a trial large enough to find out. The authors graded the certainty of the evidence as low. The next move is a trial powered for outcomes, not glucose readings.